mlx-bun 0.2.0 → 0.4.0

package/docs/reference/server-config.md

@@ -1,101 +1,346 @@
 # Server configuration
 
-Every start flag for `mlx-bun serve` (and `mlx-bun pi`, which shares the
-same flag set), what each does, its default, and — the part that bites —
-**which combinations compose and which don't**. For the request/response
-wire format (chat/messages/responses bodies, SSE grammar, tool calling,
-adapters admin) see [server-api.md](./server-api.md); this doc is about
-*starting* the server.
-
-The flags are parsed in `serverRuntimeFlags()`
-([src/cli.ts](../../src/cli.ts)); the runtime knobs live in
-`ServerOptions` ([src/server.ts](../../src/server.ts)). The serial and
-continuous scheduling mechanisms and their declaration are in
-[src/serve/generation-gateway.ts](../../src/serve/generation-gateway.ts);
-the design rationale is in
-[docs/design/parallel-slots.md](../design/parallel-slots.md).
+The canonical reference for **every start flag, every `MLX_BUN_*`
+environment variable, and every serving feature** of `mlx-bun serve` (and
+`mlx-bun pi`, which shares the flag set). Each default here is read from
+the code, not from memory: flags are parsed by `serverRuntimeFlags()` /
+`applyDecodeRoute()` in [src/cli.ts](../../src/cli.ts) (the `SERVER_FLAGS`
+help block is the same list), runtime knobs land in `ServerOptions`
+([src/server.ts](../../src/server.ts)), environment reads go through
+[src/runtime-config.ts](../../src/runtime-config.ts), and the `--isolate`
+proxy lives in [src/serve/isolate.ts](../../src/serve/isolate.ts). For the
+request/response wire format see [server-api.md](./server-api.md); for
+CLI verbs see [cli.md](./cli.md).
+
+The server runs generation through method-neutral sessions. A model binding
+supplies execution planning, its supported methods, media preparation and
+diagnostics. Stop/tool parsing runs directly in session delivery, so it can stop
+generation before another token is produced. This adds no architecture flag.
+
+Sections: [Start flags](#start-flags) · [`--isolate` semantics](#--isolate-semantics)
+· [Per-request overrides](#per-request-overrides) · [Environment variables](#environment-variables)
+· [Execution modes](#execution-modes-serial-vs---batch-n) · [Compatibility matrix](#compatibility-matrix)
+· [Fidelity tiers](#fidelity-tiers-and-the-decode-route---l1----l2) · [Feature matrix](#feature-matrix)
+· [Performance & recipes](#performance-characteristics--recipes) · [`GET /stats`](#observability--get-stats).
 
 ## Start flags
 
-| Flag | Arg | Default | What it does |
-| --- | --- | --- | --- |
-| `--host` | addr | `127.0.0.1` | Interface to bind. Loopback-only by default (mlx_lm.server parity); pass `--host 0.0.0.0` to expose the server on your network. |
-| `--port` | n | `8080` | Listen port. A pre-flight probe refuses to start if the port is already serving. |
-| `--memory-budget` | GB | machine RAM × 0.75; GLM-5.2: min(25 GiB, physical RAM) | Admission ceiling. A request whose prompt fits is always admitted: a broad client `max_tokens` is capped to the room remaining under the safe context (`max_tokens` is a ceiling, not a promise), so agent clients with family-wide defaults never 400 over a small overshoot. Only a prompt that leaves no generation slot is rejected with **400** (`type: memory_admission`) — the alternative is an uncatchable GPU OOM. The safe-context ceiling bills the active `--kv-quant` scheme at its true bytes/element (a 4-bit cache admits the ~3.5× larger window it actually enables; `turbo` is still billed bf16, conservatively). Admission never refuses to *start*: a budget too small for any context logs a warning and serves anyway (requests then get the per-request 400). GLM-5.2 additionally runs its exact process equation before opening resident weights or expert slabs and caps MLX to the plan's allocator reserve. **Decimal GB (×10⁹).** |
-| `--prompt-cache` | GB | `8` | RAM prefix-KV cache (byte-capped LRU; `0` disables). **PREFIX SHARING (2026-07-05): serves are NON-CONSUMING** — a hit hands out zero-copy clones and leaves the donor entry intact, so N agents (or a brand-new session) sharing a system prompt all reuse ONE prefill without destroying each other's conversation entries (the old consume-and-trim take cannibalized the donor). Extended entries supersede their prefix-ancestors (when trimmable), so a conversation stays one entry. Works on BOTH lanes and tiers over the SSD store (Layer 0). |
-| `--isolate` | (bool) | off | Runtime isolation ([runtime-isolation.md](../design/runtime-isolation.md)): the inference engine runs as a CHILD process on a unix socket; this process stays a pure UI/API reverse proxy — instant under any GPU load, survives engine crashes (auto-respawn, in-flight requests get a clean 502). Measured cost ≈0 (−0.4% tok/s, +2 ms TTFT, per-token SSE granularity preserved). `/ws/chat` is not proxied yet (501). `/engine` reports child pid/restarts. |
-| `--model-pool` | n | `1` | With `--isolate`: max RESIDENT model engines. Requests naming another model (**exact `/v1/models` id** — fuzzy strings keep mlx-lm's ignored-field semantics) spawn/route to that model's own engine child (SPAWN-OVERLAP: the new model loads while the old one keeps serving; nobody's stream is interrupted). Over the cap, the LRU engine drains, **demotes its prompt cache to the SSD tier**, and exits — switching back respawns it with state restored from disk. Measured (M1 Max, cpm5⇄qwen0.8b, pool 1): switch 1.5 s, switch-back 1.2 s with `cached_tokens` 103/104 — the conversation survived the eviction. The pool drives this eviction via `POST /admin/drain` — unix-socket-only (never exposed on the public TCP listener) — which quiesces the gateway and demotes the whole prompt cache to the SSD tier before eviction, responding `{ drained: true, demotions: N }`. |
-| `--ssd-cache` | dir | off | SSD cold tier under the prompt cache ([docs/design/ssd-kv-cold-tier.md](../design/ssd-kv-cold-tier.md)): prefix KV spills to disk on RAM eviction AND idle demotion, is snapshotted after requests settle (debounced 1 s; the flush is idle-gated per tensor so it never taxes an active decode — `MLX_BUN_SSD_WRITEBEHIND=0` disables), and **survives restarts** — a long-context agent re-attach restores via a bounded streamed copy instead of re-prefilling (measured 13.7k-token prefix: 12 s → ~0.25 s TTFT; 0% decode overhead; restore transient = live entry + one tensor, no mapping outlives the load). Dirty snapshots remain tracked until the atomic SSD store confirms them; `SIGINT`/`SIGTERM` drains active requests and flushes them before exit, and `POST /admin/cache/flush` exposes the same explicit boundary. Entries are keyed by model fingerprint **+ effective kv scheme** + tokenizer hash + adapter ns; incompatible/corrupt files self-quarantine. **Layer 0 (2026-07-05): the tiering lives inside PromptCache.take(), so BOTH lanes restore from disk — batch-scheduler joiners included — and every put (batch rows too) schedules the write-behind snapshot.** Requires the RAM cache. |
-| `--ssd-cache-max` | GB | `32` | SSD tier byte cap (binary GiB); oldest-mtime entries evicted at the cap. |
-| `--ssd-cache-verify` | (bool) | off | Verify every tensor hash on restore before it is copied in (roughly doubles restore reads) — integrity paranoia only; the header hash is always verified. |
-| `--ssd-demote-idle` | sec | `300` (with `--ssd-cache`) | Idle demotion (Layer 0): prompt-cache entries unused this long spill to the SSD tier and **free their GPU memory** — unified memory drains between agent bursts while every prefix stays reachable (the next hit restores via the bounded streamed copy, ~0.25 s for a 13.7k-token entry vs a 12 s re-prefill). Swept only when the engine is fully idle (never drains a running batch). `0` disables. |
-| `--batch` | n | `8` | Max concurrent requests decoded together. **Default flipped 1→8 (2026-07-05)** after GATE-B1-SPEED: a LONE request through the unified engine IS the serial engine — adopted serial-class caches, compiled decode, prompt cache + SSD restore, byte-identical output, 0.992–0.996 paired decode ratios — so the cap only changes behavior when concurrent requests actually arrive (the agentic sub-agent workload: 4–8 coding agents against one local server). `--batch 1` pins strict serial (arrival-independent numerics). See [Execution modes](#execution-modes-serial-vs---batch-n). `--decode-concurrency` is accepted for drop-in compatibility (mlx_lm.server's cap, default 32). |
-| `--kv-quant` | `config`\|`off`\|`4`\|`8`\|`turbo[:k<bits>v<bits>]` | `off` (bf16) | KV-cache quantization. **Default flipped to bf16 2026-07-05** (naked = L1): quantized KV measured 5–20% slower decode than bf16 at ≤16k on every model — on mlx-lm too — so it pays only in **memory headroom** (e.g. ~1.3 GB on the 12B @16k) and is an explicit opt-in (`--kv-quant …` or `--l2`, whose preset passes `config`). `off` = bf16 (L1). `4`/`8` = **uniform** bits (group 64, start 0) — the scheme mlx-lm exposes as `--kv-bits`/`--kv-group-size`/`--quantized-kv-start`. In the L1 kernel set (`--l1 --kv-quant 8` = fused-sdpa off) our unfused quantized SDPA is **op-for-op identical** to mlx-lm's `quantized_scaled_dot_product_attention`, so uniform-quant is **bit-exact L1** (the fused-sdpa-ON path is optiq-aligned). `config` = **per-layer mixed-precision** from `kv_config.json` — optiq-only, no mlx-lm analog → **L2**. `turbo[:k<bits>v<bits>]` = **TurboQuant** ([docs/design/turboquant-kv.md](../design/turboquant-kv.md)): rotation-based KV quantization (asymmetric-affine int8/int4/int2/int5 keys, FWHT + Lloyd-Max values) — default `k8v3` (2.56× compression at head_dim 128); `kBits` ∈ {2,4,5,8}, `vBits` ∈ {2,3,4,5,8}. A separate axis from the scheme above (mutually exclusive with `config`/`4`/`8`); v1 is dequantize-on-fetch (stock `ops.sdpa`, no fused kernel) and **full-attention layers only** — sliding-window layers stay bf16 (one-time warning, never a throw). Under `--batch N`, a per-layer `config` scheme **batches** — full-attention layers since Phase 3.1 (MiniCPM5), rotating/sliding layers since milestone 2 (gemma's whole kv_config), so every shipped `kv_config.json` now batches. The scheduler applies the mixed scheme per row, gated bit-exact per row vs the serial composition (`tests/batched-kv-quant-parity.test.ts` — compositions no other stack ships). Uniform `4`/`8` still routes those requests serial (quantizedKvStart threshold semantics). **`turbo` is solo-only, unconditionally** — `TurboQuantKVCache` is a novel `Cache` implementation (no merge/filter/temporalView), so it's excluded from the batch scheduler by construction; `GenerationGateway.place` also refuses continuous scheduling explicitly (belt + braces — both layers exist on purpose). |
-| `--paged-kv` | (bool) | off | **OPTIONAL vLLM-style paged KV cache** ([docs/design/paged-kv-cache.md](../design/paged-kv-cache.md)): full-attention layers store K/V in fixed-size **block pools** (host-side block table, `ops.takeAxis` gather back to contiguous before the stock SDPA — no new attention math, no fused kernel in v1). Env: `MLX_BUN_PAGED_KV=1`. **Gated bit-exact vs the plain path** (`tests/paged-kv-parity.test.ts`: identical single-forward logits + identical 48-token greedy trajectory on the 12B). **v1 scope: serial batch=1, Gemma4-family, bf16** — with no explicit `--batch` the CLI pins `--batch 1`; the server **refuses** (fails startup, never silently downgrades) `--batch N>1`, any `--kv-quant`, and `--draft-model`; sliding-window layers keep the rotating scheme (mixed paged-full + rotating-sliding is the supported shape). Media (vision/audio) and LoRA-adapter **requests** are scoped OUT per request — they run the plain cache path (v1 non-goal cells), never a 400. Paged requests **bypass the prompt cache** (fresh caches per request; `--ssd-cache` sees nothing) and run **uncompiled decode** (a data-dependent block-list length can't shapeless-replay). Expect a small decode cost at batch=1 — the per-step gather copy buys nothing until the batched/CoW follow-ups; this ships the correctness-proven abstraction, honestly labeled. Pool sized from prompt+`max_tokens` at request setup; exhaustion is a typed error (accounting tripwire), never truncation. |
-| `--paged-kv-block-size` | n | `256` | Tokens per KV block (`--paged-kv` only). 256 = the plain cache's growth step, so v1's allocation granularity is a permutation of today's into reusable slots, not a new tuning axis. |
-| `--adapter` | dir | none | Mount a LoRA adapter at startup (same machinery as `POST /v1/adapters`; the adapter id is the directory's basename) and make it the **default** for requests that send no `adapter` field. A request's explicit `adapter` — including `"none"` — always wins, and hot-swap via `/v1/adapters` is unchanged. `--adapter-path` is accepted as the mlx_lm.server-named alias. A bad adapter fails startup loudly rather than silently serving the base model. This is the flag `mlx-bun train`'s completion message points at. |
-| `--draft-model` | path/query | none | **Speculative decoding**: a drafter proposes tokens the main model verifies in one forward — exact results, faster decode when drafts land. Resolves like the main model. The artifact's **kind is auto-detected**: a full same-tokenizer model (mlx_lm.server parity — L1: token-for-token vs mlx-lm's spec path; tokenizer-family mismatch fails startup, upstream silently accepts ~0%), a Gemma `-assistant` KV-borrowing drafter (L2 vs optiq spec_generate; ~1.09× γ=1 on 12B, no training), a locally-trained **DSpark** checkpoint (`dspark.json`), or one of **DeepSeek's released DSpark drafters** (DeepSpec `Gemma4DSparkModel` config stamp, e.g. `deepseek-ai/dspark_gemma4_12b_block7` for the 12B — oracle: DeepSpec's reference at temperature 0, token-for-token). All four share ONE serve loop. A confidence-calibrated DSpark checkpoint additionally self-schedules its draft length per round (threshold pruning — fewer wasted verify positions; uncalibrated checkpoints draft fixed-length). Mounting a draft routes **every** request to the serial lane (upstream `is_batchable = draft is None`). Pays on slow targets (12B+); fast small models lose to the draft overhead. Composes with structured output (the constrained verify walk). Prompt-cache reuse is bypassed on the spec path (v1). Telemetry: `usage.speculation` (`drafted`/`accepted`/`targetCalls`). |
-| `--draft-kind` | `two-model`\|`assistant`\|`dspark`\|`deepspec`\|`mtp`\|`ngram` | auto | Override the draft-artifact kind detection (`dspark.json` → dspark; `Gemma4DSparkModel` architecture stamp → deepspec; `*_assistant` config → assistant; `*_mtp` model_type → mtp, the target-release-native multi-token-prediction head — it shares the target's embeddings/lm-head, defaults `--num-draft-tokens` to its trained `block_size - 1`, and partial rejects roll the target's recurrent DeltaNet caches back via bit-exact snapshot/replay; else two-model). `mtp` alone (no `--draft-model`) mounts the companion bundled at `<model>/mtp/` when present — single-repo artifacts — and refuses otherwise. **`ngram` is never auto-detected — it has no artifact**: MODEL-FREE prompt lookup (drafts copied from the request's own prompt+generation when its trailing k-gram occurred earlier in the stream; port of prompt-lookup decoding / vLLM's `ngram` proposer). Mount it ALONE — `--draft-kind ngram` with a `--draft-model` is refused, as is any other kind without one. Zero weights/memory, lossless by the same verify (gated: serve-loop output token-identical to non-spec greedy on e4b, tests/spec-ngram.test.ts); a no-match round degrades to one plain target step. Best on agentic/RAG/code-edit traffic that re-emits context spans. |
-| `--num-draft-tokens` | n | `3` (`ngram`: `10`) | Drafts per verify round (mlx_lm.server's default; `mlx_lm.generate`'s is 2). A DSpark draft pins this to its trained block width (`cfg.gamma`); ngram drafting is free so its default is the reference implementation's 10. |
-| `--ngram-max` / `--ngram-min` | k | `3` / `1` | `--draft-kind ngram` only: longest/shortest trailing k-gram searched for a match (the reference values — longest first, first occurrence wins). |
-| `--mtp` | `on`\|`off` | on for GLM-5.2 | Mount the checkpoint-native GLM MTP row as the server's drafter. It uses the bounded auxiliary expert tier and exact serial verify loop (`usage.lane: "serial+spec"`). `off` removes the draft so ordinary GLM requests can use continuous batching; per-row batched MTP is post-release. Mutually exclusive with an explicit `--draft-model`/`--draft-kind`. |
-| `--context-length` | tokens | `4096` for GLM-5.2 | Context reserved by GLM's header-only resource equation. The same value becomes the request-admission ceiling and is reported in `/stats.glm52`; an impossible plan fails before committing model memory. |
-| `--thinking` | `true`\|`false` | model's own (false for CPM and GLM-5.2) | Server-wide default for the chat template's `enable_thinking` (MiniCPM5/CPM, Qwen3.5, and GLM-5.2). Full precedence: an explicit request `chat_template_kwargs.enable_thinking` always wins; else `reasoning_effort` (`"none"` → off, any other level → on) if the request sent one; else this server default; else the model's own default. GLM renders disabled thinking as `<think></think>` and enabled thinking as an open `<think>` generation primer plus `Reasoning Effort: Max`. |
-| `--temperature` | n ∈ [0,5] | `generation_config.json` | Server-wide sampling default. Per-request `temperature` still wins; the browser chat (sends none) inherits this. `--temp` is accepted as an alias (mlx_lm.server compat); explicit `--temperature` wins if both are given. **Migration note:** mlx_lm.server's `--temp` *default* is `0.0` (unset-temperature requests are greedy there); mlx-bun falls back to the model's `generation_config.json`, then `0.7` — pass `--temp 0` for mlx-lm's behavior. |
-| `--top-p` | n ∈ [0,1] | `generation_config.json` | Server-wide top-p default (per-request `top_p` wins). |
-| `--top-k` | n ∈ [0,1e6] | `generation_config.json` | Server-wide top-k default (per-request `top_k` wins). |
-| `--max-tokens` | n | GLM-5.2: `128`; otherwise `65536` chat / `512` raw completion | Completion cap when a request omits `max_tokens` (mlx_lm.server flag; its default there is 512 — `--max-tokens 512` reproduces mlx_lm.server exactly). For GLM this is also reserved by the pre-open resource equation and must fit inside `--context-length`. |
-| `--allow-private-media` | (bool) | off | Let `image_url`/`audio_url` content parts fetch from **private/loopback/link-local** hosts (a NAS, another LAN box). Off by default: a request's URL is attacker-controlled input, so remote media fetches (`src/media-fetch.ts`) refuse non-http(s) schemes and destinations in private/loopback/link-local/CGNAT ranges — including hosts that *resolve* there and every **redirect hop** (each hop re-passes the whole policy) — which keeps the server from being steered at cloud metadata (`169.254.169.254`) or LAN-internal services (SSRF). Independent of the flag, every remote fetch has a **10 s wall-clock timeout** and a **64 MB response cap** (streaming-enforced, not just Content-Length); violations surface as clean `400`s (`prompt build failed: …`). `data:` URLs decode locally and are never policy-checked. |
-| `--no-open` | (bool) | off | Skip the automatic browser open on start. By default an interactive terminal session opens `http://<host>:<port>/#/chat` once the server is ready; pass this flag to suppress it (e.g. headless or non-TTY environments already skip it). |
-| `--hlg-sampling` | `on`\|`off` | off | Piecewise tone-curve (HLG) sampling: rolls off the top-token region, boosts the mids, gentles the tail. The overall gain folds from `--temperature`. See [docs/archive/hlg-sampling.md](../design/hlg-sampling.md). |
-| `--hlg-width` | nats | `4` | HLG mid-region half-width (nats). Only meaningful with `--hlg-sampling on`. |
-| `--hlg-shoulder` | nats | `4` | HLG highlight rolloff scale (nats). Only meaningful with `--hlg-sampling on`. |
-| `--hlg-toe` | nats | `6` | HLG shadow rolloff scale (nats). Only meaningful with `--hlg-sampling on`. |
-| `--hlg-pivot-offset` | nats | `6` | HLG pivot point: nats below the top token. Only meaningful with `--hlg-sampling on`. |
-| `--expert-offload` | (bool) | off | **MoE models only.** Serve experts from a page-aligned file mmap (built on first use). Keeps the model out of memory pressure — physical footprint ≈ active params. Ignored with a warning on dense models. Bit-exact with the resident path. |
-| `--l1` | (bool) | — | **Parity tier ALIAS:** bit-for-bit IDENTICAL to mlx-lm (bf16 KV, compiled decode + compiled activations — the faithful `@mx.compile` geglu/swiglu). **No tier given ⇒ `--l1`** (decided 2026-07-05: the L1 faithful kernel set matches mlx-lm 1.00× on every model and no output-changing lever has beaten that baseline in a paired A/B — each is opt-in until it does). Expands to the fastest set of per-fork flags that still holds the guarantee; any explicit per-fork flag (`--kv-quant`/`--fused-sdpa`/`--compiled-decode`/`--compiled-activations`) overrides one. See [docs/design/faithful-l1-consolidation.md](../design/faithful-l1-consolidation.md) and [parity-tier-dag.md](../design/parity-tier-dag.md). |
-| `--l2` | (bool) | — | **Parity tier preset:** bit-for-bit IDENTICAL to mlx-optiq (quantized KV per `kv_config.json` + fused N-tiled prefill SDPA + **stock unfused decode** — the composition the optiq goldens track, `scripts/regen-kvq-goldens.ts`). |
-| `--compiled-decode` | on\|off | on | Replay the per-step decode graph in C++ (`MLX_BUN_COMPILED_DECODE`). Bit-exact A/B lever. **Serial lane only** (see note below). **Gemma4-dense only** — LoRA, MoE, and non-Gemma4 models (MiniCPM5 / Qwen3.5) run eager; an unsupported step falls back to eager for the rest of that generation. |
-| `--compiled-activations` | on\|off | on | Route the geglu/swiglu activation through mlx-lm's `@mx.compile` closure (`MLX_BUN_COMPILED_GEGLU` + `MLX_BUN_COMPILED_SWIGLU`) — the **faithful** kernel: same libmlx graph as mlx-lm → **bit-exact** AND one dispatch instead of ~9. `off` = the uncompiled composition (same L1 parity, slower). Toggles gemma geglu + MiniCPM5 swiglu; qwen3/qwen3.5/universal compile unconditionally. |
-| `--fused-sdpa` | on\|off | follows `--kv-quant` | Fused SDPA path for quantized prefill/continuation (inverted env `MLX_BUN_NO_FUSED_SDPA`). Defaults to the composition its oracle uses: **on** under `--kv-quant config` (the optiq-golden composition), **off** under uniform `4`/`8` (mlx-lm's `quantized_scaled_dot_product_attention`, the bit-exact L1-eligible scheme) and bf16 (no-op there). **Serial lane only.** |
-| `--force-wire` | (bool) | off | Wire weights into memory for the whole generation (`MLX_BUN_FORCE_WIRE`). Near-ceiling models (e.g. 26B) need it. **Serial lane only.** |
+How flags are parsed (`src/cli.ts` `opt`/`flag`/`positional`):
+
+- A value flag is `--name value` (space-separated; `--name=value` is not
+  recognized). A value flag with no following token takes its default.
+- **Unknown flags are silently ignored** — there is no unknown-flag error.
+  mlx_lm.server's `--kv-group-size` and `--quantized-kv-start` are in this
+  class: `src/serve/isolate.ts` lists them as value-taking so the engine
+  argv builder does not mistake their value for a model positional, but
+  `serve` never reads them (uniform `--kv-quant 4|8` is fixed at group
+  size 64, start 0).
+- Aliases: `--adapter-path` = `--adapter`; `--temp` = `--temperature`
+  (explicit `--temperature` wins); `--decode-concurrency` = `--batch`
+  (`--batch` wins); `--query`/`-q` = the positional model query.
+- `mlx-bun pi` consumes the same flag set; when it attaches to an
+  already-running server every server-shaping flag except `--port` is
+  reported as ignored (the running server keeps its configuration).
+
+**Lane/tier column:** `serial` = the strict single-queue executor;
+`batch` = the continuous scheduler (`--batch N`); `both` = lane-independent.
+L1 = bit-exact vs mlx-lm, L2 = bit-exact vs mlx-optiq, Lab = no external
+oracle. See [Fidelity tiers](#fidelity-tiers-and-the-decode-route---l1----l2).
+
+### Model selection
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `[query]` (positional) | registry query | auto-pick | both | First non-flag token after `serve`: a registry query (`"12B"`, `"e4b"`, a repo substring). Omitted: serves e4b when downloaded, else the largest downloaded model that fits; a fresh install downloads a small starter model first, then e4b in the background. |
+| `--model` | path \| query | none | both | mlx_lm.server's spelling, and the **highest-precedence** selector: a directory containing `config.json` loads straight from that path (`scanSnapshot`, no registry lookup — the bench/CI shape; an HF-cache path recovers its repo id from `models--org--name/snapshots/`); anything else resolves like the positional query. Precedence: `--model` > positional > `--query` > auto-pick. Under `--isolate`, the parent resolves the model and pins `--model <path>` on the engine child. |
+| `--query` / `-q` | registry query | none | both | Same as the positional query (lowest-precedence explicit selector). `-q` is the `mlx-bun pi` spelling. |
+
+### Network
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--host` | addr | `127.0.0.1` | both | Interface to bind. Loopback-only by default (mlx_lm.server parity); `--host 0.0.0.0` exposes the server on your network. The browser-open uses `localhost` when the bind is `0.0.0.0`/`::`. |
+| `--port` | n | `8080` | both | Listen port. Before loading weights the CLI probes `http://localhost:<port>/v1` and refuses to start if something is already serving there (skipped under `--unix`). |
+| `--unix` | socket path | none | both | **Internal.** Listen on a unix domain socket instead of TCP — the engine half of `--isolate` (`ServerOptions.unixSocket`; `hostname`/port are ignored, stale socket files are unlinked before bind). Usable directly for socket-level integration. |
+
+### Memory, admission, and caches
+
+Memory estimates are advisory by default. The server attempts requests without
+rejecting a prompt or shortening `max_tokens` based on predicted GPU headroom.
+`fit` and `/stats.admission.max_safe_context` remain estimates; they do not
+set the default execution limit. Explicit `--memory-budget`, `--kv-budget`,
+and context caps remain opt-in restrictions. GLM-5.2's allocated context
+layout retains its configured capacity.
+
+When batch rows finish, full-attention caches drop padding shared by every
+surviving row, matching MLX-LM while preserving each row's absolute position.
+
+Qwen prefills longer than four tokens materialize each layer and its cache
+state before continuing, bounding deferred buffer lifetimes. Execution has
+no reservation checks between layers or during decode. Serial requests
+finish outstanding SSD snapshots before generation. The RAM prefix-cache
+cap controls retained entries; it does not promise that every workload fits.
+Actual allocation failures use the ordinary error path when recoverable;
+a native Metal allocation failure can terminate the process.
+With `--ssd-cache`, an exact recurrent-cache hit persists its boundary before
+transferring the live state to the request. This keeps that boundary reusable
+after the recurrent state advances. Already-durable boundaries need no write;
+a failed required store retains the cache entry and refuses the transfer.
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--memory-budget` | GB (decimal, ×10⁹) | unset; memory estimates are advisory | both (per request) | Explicit opt-in budget. Sets the allocator limit and enforces the fit estimate: a prompt leaving no generation slot gets **400** (`type: memory_admission`); a broader completion cap is reduced to the remaining context. Estimates account for active KV quantization and prefill temporaries. Not an aggregate batch cap — see `--kv-budget`. GLM-5.2 also uses this value in its resource plan before loading. |
+| `--kv-budget` | GB (decimal) | off | batch | Aggregate KV budget across concurrently admitted batch rows: a joiner whose projected KV (prompt + `max_tokens`, window-capped) would exceed it **queues** until rows finish; a request over the budget alone is rejected. Without it, N large-context rows can collectively exceed memory. Reported in `/stats.batch.kv_budget_bytes`. |
+| `--prompt-cache` | GB (binary GiB) | `8` | both | RAM prefix-KV cache (byte-capped LRU). `--prompt-cache 0` disables it. RAM hits lend zero-copy views and leave the donor intact, including recurrent Qwen and method state. SSD persistence runs in the background; RAM reuse never waits for a write. Explicit cache flush and graceful shutdown wait for durability. Batch-lane joiners `take()` the longest usable prefix at admission; a row that finishes without ever merging `put()`s its caches back (merged rows' entries age out). |
+| `--ssd-cache` | dir | off | both | SSD cold tier under the prompt cache ([docs/design/kv-cache.md](../design/kv-cache.md)): prefix KV spills to disk on RAM eviction and idle demotion, is snapshotted after requests settle (debounced, each tensor step owns the generation lock so it cannot overlap inference; `MLX_BUN_SSD_WRITEBEHIND=0` disables), and **survives restarts**. Entries are keyed by model fingerprint + effective KV scheme + backend numerical identity + tokenizer hash + adapter namespace. MLX includes its runtime version and GPU architecture; older numerical identities use separate directories. `SIGINT`/`SIGTERM` drains active requests and flushes dirty snapshots before exit (`MLX_BUN_SHUTDOWN_TIMEOUT_MS`); `POST /admin/cache/flush` is the explicit boundary. Requires the RAM cache. |
+| `--ssd-cache-max` | GB (binary GiB; `0` = unlimited) | unlimited | both | Optional SSD tier byte cap; oldest-mtime entries are evicted only when a positive cap is configured. Warns and is ignored without `--ssd-cache`. |
+| `--ssd-cache-verify` | (bool) | off | both | Verify every tensor hash on restore (reads all bytes eagerly, defeating lazy fault-in) — integrity paranoia only; the header hash is always verified. Warns and is ignored without `--ssd-cache`. |
+| `--ssd-demote-idle` | seconds | `300` with `--ssd-cache`, else off | both | Prompt-cache entries unused this long spill to the SSD tier and free their GPU memory; the next hit restores them. Swept only when the engine is fully idle. `0` disables. Warns and is ignored without `--ssd-cache`. |
+| `--generation-checkpoint` | output tokens | off | qualified requests | Queues an owned in-flight generation snapshot every N emitted tokens. Repeating the identical request after a restart replays the saved assistant prefix and continues from its already-sampled next token. Requires `--ssd-cache`. Shared execution qualifies ordinary requests with supported row-cache layouts, including Qwen, Llama, MiniCPM and Gemma4, with or without adapters, excluding media, grammar, fill, paging, logprobs or speculative methods; existing serial qualification remains available. Explicit cache flush drains queued writes. Completed generations remove their checkpoint, while client cancellation preserves the latest checkpoint. |
+
+Generation checkpoint identity includes stop strings, sampling/KV policy, the
+resolved execution method, compiled/grammar policy, and artifact/implementation
+identity. Changing any of these starts a new completion. Version-2/3 generation keys are not
+selected for automatic resume. Prefix-cache files retain their state codec
+format, but automatic reuse also requires the model binding's numerical
+identity. MLX runtime or GPU architecture changes select a fresh cache
+directory and leave the previous directory intact. A different backend codec
+identity refuses incompatible files. Adapter namespaces include the mounted weight contents and
+scale, so a changed adapter starts with a fresh prefix and checkpoint.
+Native media and grammar preparation waits for the generation execution lease
+and retains a reservation through completion. At most one media preparation and
+`--batch` grammar preparations can be retained. Generation and preparation
+queues each allow 64 waiting requests; overflow returns `429` before a response
+stream opens, or a terminal stream error after it opens. Disconnects release
+queued reservations and completed preparation resources.
+
+Ordinary requests in the continuous scheduler can join an ongoing prefill
+between chunks when their execution context is compatible and batch capacity
+is available. Scheduling also budgets the initial prefill work in a cohort, so
+long prompts can be prepared one at a time through the continuous executor
+while decode remains batched. Subsequent forwards process real-token chunks from admitted requests
+together. Each request keeps its own cache-maintenance and checkpoint boundaries.
+Arrival timing can change batch geometry and near-tie generated text. Qwen MTP,
+prompt lookup and standalone drafting use the same target prefill driver.
+Their providers prepare companion state through a separate interface;
+scheduling determines admission and work budgets.
+
+Disconnected serial requests leave the admission queue immediately. Active AR
+requests check cancellation between prefill chunks and decode steps; a native
+operation already running completes before that boundary.
+
+### Runtime isolation
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--isolate` | (bool) | off | both | Run the inference engine as a **child process** on a unix socket while this process owns CPU application routes and proxies inference — instant under any GPU load, survives engine crashes (auto-respawn). Full semantics in [`--isolate` semantics](#--isolate-semantics). |
+| `--model-pool` | n (≥1) | `1` | both | With `--isolate`: max **resident** model engines. A request whose `model` field is an **exact** `/v1/models` id spawns/routes to that model's own engine child (spawn-overlap: the new model loads while the old keeps serving). Over the cap the least-recently-used engine is drained (`POST /admin/drain` over its socket), demotes its prompt cache to the SSD tier, and exits; switching back respawns it with state restored from disk. Without `--isolate` the flag warns and is ignored. |
+
+### Scheduling
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--batch` | n (integer ≥ 1) | `8` | batch / L1 | Max concurrent requests decoded together. A lone admitted request uses the B=1 fast path (adopted serial-class caches, compiled decode, prompt cache + SSD restore — byte-identical to serial); a batch layout only exists once a second request arrives. `--batch 1` pins the strict serial executor (arrival-independent numerics — use it for golden regeneration). `--decode-concurrency` is accepted as the mlx_lm.server alias, but its semantics differ there (per-`BatchGenerator` decode parallelism, default 32). See [Execution modes](#execution-modes-serial-vs---batch-n). |
+
+### KV cache
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--kv-quant` | `config` \| `off` \| `4` \| `8` \| `turbo[:k<bits>v<bits>]` | `off` (bf16) | see row | KV-cache quantization. `off` = bf16 (the L1 default: quantized KV measured 5–20% slower decode than bf16 at ≤16k on both stacks, so it is an opt-in that buys **memory headroom**). `4`/`8` = **uniform** bits (group 64, start 0 — mlx-lm's `--kv-bits` scheme); with fused-sdpa off (the default for uniform) our unfused quantized SDPA is op-for-op mlx-lm's `quantized_scaled_dot_product_attention`, so uniform is **bit-exact L1**; uniform requests use shared B=1/B>1 execution where the loaded cache layout supports affine conversion. `config` = **per-layer mixed precision** from the model's `kv_config.json` — optiq-only → **L2**; per-layer configs **batch** on every shipped model (full-attention and rotating layers; gated bit-exact per row vs the serial composition). `turbo[:k<bits>v<bits>]` = **TurboQuant** ([docs/design/turboquant.md](../design/turboquant.md)): rotation-based KV quantization, default `k8v3`; `kBits` ∈ {2,4,5,8}, `vBits` ∈ {2,3,4,5,8}; a separate axis (mutually exclusive with `config`/`4`/`8`), uses the existing decoded or deferred rotated-value attention path, **full-attention layers only** (sliding-window layers stay bf16 with a one-time warning), refuses head dims outside `TURBOQUANT_HEAD_DIMS`, and supports shared ordinary and grouped drafting execution at B=1/B>1 including positive library conversion thresholds when the cache layout qualifies. Any explicit `--kv-quant` overrides the tier preset. |
+| `--paged-kv` | (bool) | off | shared / gated bit-exact | **Optional vLLM-style paged KV cache** ([docs/design/kv-cache.md](../design/kv-cache.md)): full-attention layers store K/V in fixed-size block pools (host-side block table, gather back to contiguous before the stock SDPA; no new attention math). Env mirror `MLX_BUN_PAGED_KV=1`. Paging retains the configured batch size; request state owns independent block pools through the shared cache-layout interface. `createServer` **refuses** any `--kv-quant` (incl. turbo), `--draft-model`, and non-Gemma4 models. Media and adapter requests run the plain cache path per request. Paged requests bypass the prompt cache and run uncompiled decode; pool exhaustion is a typed error, never truncation. Gated bit-exact vs the plain path (`tests/parity/paged-kv-parity.test.ts`). |
+| `--paged-kv-block-size` | n | `256` | shared | Tokens per KV block (`--paged-kv` only); 256 = the plain cache's growth step. |
+
+### Adapters and speculative decoding
+
+In the current unreleased source, Qwen MTP publishes aligned target/draft
+prefill snapshots through the shared prompt cache by default. The RAM budget
+includes the draft KV and pending hidden row; configured SSD persistence
+queues those tensors with the target state and supports reuse after restart.
+RAM hits retain immutable snapshots for later requests. Target, draft-weight,
+adapter-revision and KV-policy identities separate incompatible state.
+`usage.cached_tokens` reports the reused target prefix.
+`MLX_BUN_MTP_PROMPT_CACHE=0` disables this reuse. Generated-output snapshots
+remain unfinished.
+
+In the current unreleased source, Qwen uniform affine KV4 with
+`quantizedKvStart=0` supports configured speculative execution by default.
+`MLX_BUN_QWEN_SPEC_KV4=0` restores the ordinary-decode compatibility control. Recurrent state and draft KV keep their original precision. Shared Qwen MTP
+also supports bf16, uniform KV8 and TurboQuant at start zero; per-layer mixed
+KV remains incompatible. Strict legacy serial speculation retains bf16/KV4.
+The KV4 control does not disable bf16, KV8 or TQ shared speculation. Six paired M4 Pro combined suites and two completed Kanban tasks support
+this default. Configuring a drafter and selecting KV4 remain explicit choices.
+
+For controlled experiments, `MLX_BUN_RD_PREFILL_CHUNK` sets the prefill chunk,
+default 2048. Shared ordinary and speculative methods use the group's captured
+default; an explicit group option overrides the environment and a library
+request's `prefillChunkSize` overrides that default for its request.
+`MLX_BUN_RD_CONTEXT_LIMIT` sets an explicit request context cap
+without enlarging it. Both require positive integers; the context cap is unset
+by default. These are benchmark controls, not changes to the published model
+profile or sampling policy.
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--adapter` | dir | none | compatible adapter groups | Mount a LoRA adapter at startup (same machinery as `POST /v1/adapters`; the id is the directory basename) and make it the default for requests without an `adapter` field. A request's explicit `adapter` — including `"none"` — wins; hot-swap via `/v1/adapters` is unchanged. A bad adapter fails startup. Alias `--adapter-path`. |
+| `--draft-model` | path \| query | none | shared Qwen methods; full-attention/rotating lookup/standalone and Gemma assistant/DeepSpec/DSpark; other combinations serial | **Speculative decoding**: a drafter proposes tokens the target verifies in one forward — exact results, faster decode when drafts land. Resolves like the main model. Kind is auto-detected: a full same-tokenizer model (mlx_lm.server parity, L1 token-for-token; tokenizer-family mismatch fails startup), a Gemma `-assistant` KV-borrowing drafter (L2 vs optiq `spec_generate`), a locally trained **DSpark** checkpoint (`dspark.json`), or a released DeepSpec `Gemma4DSparkModel` drafter. Qwen MTP, prompt lookup and standalone drafting use the shared executor at one or several active rows, including logprobs. Supported full-attention and rotating targets, including Llama, MiniCPM and Gemma, also share bf16, uniform KV4/KV8 or per-layer affine KV or TurboQuant prompt lookup and standalone drafting. Rotating target transactions retain different accepted lengths across the sliding window and publish paired RAM/SSD state. Gemma assistant drafting shares the same execution at B1/B>1 with bf16, uniform KV4/KV8 or per-layer affine KV or TurboQuant target storage. Its hidden-state companion and target KV use the RAM/SSD cache; donor attention owns row validity while the assistant graph stays independent of storage. DeepSpec and DSpark share bf16, uniform KV4/KV8 or per-layer affine KV or TurboQuant target execution and projected-context RAM/SSD companions. DeepSpec supports bf16 or affine-quantized draft weights; DSpark supports its Markov/RNN heads. Confidence selects each row's proposal length independently. Other provider/target combinations still run serial. Unsupported speculative compositions use ordinary execution when their own features support it. Composes with structured output; Qwen MTP, prompt lookup, standalone, Gemma assistant, DeepSpec and DSpark drafting use the shared RAM/SSD prefix cache by default (`MLX_BUN_MTP_PROMPT_CACHE=0` disables); other speculative sources start fresh. Telemetry: `usage.speculation`. |
+| `--draft-kind` | `two-model` \| `assistant` \| `dspark` \| `deepspec` \| `mtp` \| `ngram` | auto | shared Qwen methods; full-attention/rotating lookup/standalone and Gemma assistant/DeepSpec/DSpark; other combinations serial | Override drafter detection. `mtp` = a native multi-token-prediction head split from the target's release (`*_mtp` model_type; shares the target's embeddings/lm-head, defaults `--num-draft-tokens` to its trained `block_size − 1`, rolls DeltaNet caches back by snapshot/replay on partial rejects); `mtp` alone mounts the companion bundled at `<model>/mtp/` when present. Qwen MTP companions accept dense or affine-quantized projections using their checkpoint metadata; changing the draft can change acceptance and speed. `ngram` = **model-free prompt lookup** (drafts copied from the request's own prompt+generation; port of prompt-lookup decoding / vLLM's `ngram` proposer) — mount it **alone** (`--draft-kind ngram` with a `--draft-model` is refused, as is any other kind without one); lossless by the same verify, a no-match round degrades to one plain target step. Any other value fails startup. |
+| `--num-draft-tokens` | n (integer ≥ 1) | `3` (`ngram`: `10`; DSpark: pinned ≤ its trained `gamma`) | method-dependent | Drafts per verify round (mlx_lm.server's default; `mlx_lm.generate`'s is 2). |
+| `--ngram-max` / `--ngram-min` | k (integer ≥ 1) | `3` / `1` | Qwen and full-attention targets shared / other models serial | `--draft-kind ngram` only: longest/shortest trailing k-gram searched (longest first, first occurrence wins). `--ngram-min` > `--ngram-max` fails startup; either flag without `ngram` warns and is ignored. |
+| `--mtp` | `on` \| `off` | `on` for GLM-5.2 | serial+spec | GLM-5.2 checkpoint-native MTP row as the server's drafter, using the bounded auxiliary expert tier and the exact serial verify loop (`usage.lane: "serial+spec"`). `off` removes the draft so ordinary GLM requests can use continuous batching. Other model families ignore it. |
+| `--context-length` | tokens (integer ≥ 1) | `4096` (GLM-5.2) | both | Context reserved by GLM-5.2's header-only resource equation; also the request-admission ceiling, reported in `/stats.glm52`. An impossible plan fails before committing model memory. Ignored by other families. |
+
+### Sampling and template defaults
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--thinking` | `true`\|`false` (also `on`/`off`/`1`/`0`) | model's own (`false` for MiniCPM5) | both | Server-wide default for the chat template's `enable_thinking`. Precedence (`resolveEnableThinking`): explicit request `chat_template_kwargs.enable_thinking` → request `reasoning_effort` (`"none"` = off, any other level = on) → this flag → the model's default. With no explicit temperature, a no-think turn is capped at 0.7 while a think turn keeps the model's configured temperature. |
+| `--temperature` | n ∈ [0, 5] | `generation_config.json`, else `0.7` | both | Server-wide sampling default; a per-request `temperature` wins; the browser chat (sends none) inherits it. Alias `--temp` (mlx_lm.server spelling — note its *default* there is `0.0`; pass `--temp 0` for that behavior). |
+| `--top-p` | n ∈ [0, 1] | `generation_config.json`, else `0` (off) | both | Server-wide top-p default (per-request `top_p` wins). |
+| `--top-k` | n ∈ [0, 1e6] | `generation_config.json`, else `0` (off) | both | Server-wide top-k default (per-request `top_k` wins). |
+| `--max-tokens` | n ∈ [1, 1e7] | GLM-5.2: `128` (memory-plan reservation); otherwise none — an omitted cap generates until EOS/stop or an explicitly configured limit | both | Completion cap when a request omits `max_tokens` (mlx_lm.server flag). DEVIATION when unset: `mlx_lm.server` stops a defaulted request at 512 — `--max-tokens 512` reproduces it. For GLM-5.2 the value is also reserved by the pre-open resource equation and must fit inside `--context-length`. Note: `max_tokens` also feeds `--kv-budget` row projections (prompt + cap), so an uncapped request cannot fit a finite explicit KV budget; set `--max-tokens` or send a request cap when using that budget. |
+| `--hlg-sampling` | `on`\|`off` | off | serial / Lab | Piecewise tone-curve (HLG) sampling: rolls off the top-token region, boosts the mids, gentles the tail; the gain folds from `--temperature`. Design: [docs/archive/hlg-sampling.md](../archive/hlg-sampling.md). |
+| `--hlg-width` | nats ∈ [0, 100] | `4` | serial | HLG mid-region half-width. Only with `--hlg-sampling on`. |
+| `--hlg-shoulder` | nats ∈ [0, 100] | `4` | serial | HLG highlight rolloff scale. Only with `--hlg-sampling on`. |
+| `--hlg-toe` | nats ∈ [0, 100] | `6` | serial | HLG shadow rolloff scale. Only with `--hlg-sampling on`. |
+| `--hlg-pivot-offset` | nats ∈ [0, 100] | `6` | serial | HLG pivot: nats below the top token. Only with `--hlg-sampling on`. |
+
+### Media and UX
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--allow-private-media` | (bool) | off | serial (media requests) | Let `image_url`/`audio_url`/`video_url` parts fetch from **private/loopback/link-local** hosts. Off by default: a request's URL is attacker-controlled, so `src/media-fetch.ts` refuses non-http(s) schemes and private/loopback/link-local/CGNAT destinations — including hosts that *resolve* there and every redirect hop (SSRF guard). Independent of the flag, every remote fetch has a 10 s timeout and a 64 MB streaming-enforced cap; violations are clean `400`s. `data:` URLs are never policy-checked. Env mirror `MLX_BUN_ALLOW_PRIVATE_MEDIA=1`. |
+| `--no-open` | (bool) | off | both | Skip the automatic browser open. By default an interactive TTY opens `http://<host>:<port>/#/chat` once the server is ready; non-TTY runs never open. Parent-only under `--isolate`. |
+
+### Parity tier and kill switches
+
+| Flag | Arg | Default | Lane/tier | What it does |
+| --- | --- | --- | --- | --- |
+| `--l1` | (bool) | **the default** | serial route / L1 | Tier alias: bit-for-bit identical to mlx-lm — bf16 KV, compiled decode, compiled activations, fused-sdpa off. No tier flag ⇒ `--l1` (decision 2026-07-05). Any per-fork flag below overrides one preset value. |
+| `--l2` | (bool) | — | serial route / L2 | Tier alias: bit-for-bit identical to mlx-optiq — `--kv-quant config` + fused N-tiled prefill SDPA + stock unfused decode (the composition the optiq goldens track). |
+| `--l3` | — | — | — | **Removed 2026-07-05.** Passing it exits with a pointer to `--l2` / the Lab. |
+| `--compiled-decode` | `on`\|`off` | on | serial, and batch at B=1 / L1 | Replay the per-step decode graph in C++ (`MLX_BUN_COMPILED_DECODE`). Bit-exact A/B lever. Gemma4-dense only; LoRA, MoE, and other families run eager, and an unsupported step falls back to eager for the rest of that generation. |
+| `--compiled-activations` | `on`\|`off` | on | both / L1 | Route geglu/swiglu through mlx-lm's `@mx.compile` closure (`MLX_BUN_COMPILED_GEGLU` + `MLX_BUN_COMPILED_SWIGLU`). MiniCPM5 uses compiled SwiGLU for decode and prefill. `off` selects the uncompiled composition and can change prefill numerics on MLX 0.32.2. Toggles gemma geglu + MiniCPM5 swiglu; qwen3/qwen3.5/universal compile unconditionally. |
+| `--fused-sdpa` | `on`\|`off` | follows `--kv-quant`: on for `config`, off for `4`/`8` and bf16 | serial / L2 | Fused SDPA for quantized-KV prefill/continuation (inverted env `MLX_BUN_NO_FUSED_SDPA`). Defaults to the composition its oracle uses. No-op on bf16. |
+| `--force-wire` | (bool) | off | serial | Wire weights into memory for the whole generation (`MLX_BUN_FORCE_WIRE=1`). Near-ceiling models need it. The batch scheduler bypasses `generate()`'s wired scope. |
+| `--expert-offload` | (bool) | off | serial / Lab | **MoE only.** Serve experts from a page-aligned file mmap (`<model>/.mlx-bun-offload`, built on first use) — physical footprint ≈ active params. Dense models log "ignored". Bit-exact with the resident path. Env mirror: `MLX_BUN_EXPERT_OFFLOAD=<dir>`. |
 
 The default host/port (`127.0.0.1:8080`) match `mlx_lm.server`, so running
-mlx-bun alongside the Python reference server needs an explicit `--port`.
-
-The remaining levers are bit-exact A/B knobs; **the naked default is the
-L1 set** (2026-07-05 decision: an output-changing lever earns a default
-only by beating the L1 faithful baseline in a paired A/B, and none did —
-the 2026-07-05 pass measured the custom fused-decode at 1.00×, fused-gelu
-at +0–1%, the flash perf-kernel at 0.62–0.93× on e4b, and quantized KV
-5–20% *slower* decode than bf16 at ≤16k on both stacks; those losing
-kernels and their flags were **deleted** the same day — Phase 1 of
-[unified-engine-frontier-plan.md](../design/unified-engine-frontier-plan.md).
-New output-changing experiments live in the **Lab**: env-flagged, with a
-bench + expiry, until one beats the L1 baseline in a paired A/B). Flip
-the survivors to compare. They are set as `MLX_BUN_*` env vars before the
-model loads, so they apply to `mlx-bun pi` too. They affect the
-**serial** decode path (`generate()`); the batched scheduler calls
-`model.forwardHidden` directly and is unaffected by all of them except
-`--compiled-decode`, which also covers a lone batch-lane request (Phase
-3.2) — see
+mlx-bun alongside the Python reference needs an explicit `--port`.
+
+## Reproducing mlx_lm.server
+
+Numeric parity (same tokens in → same logits, bit-for-bit) is unconditional
+— no flag buys it back and none is needed. **Behavior policy** (defaults,
+caps, sampling fallbacks) is mlx-bun's own; mlx_lm.server's policy is one
+configuration of it:
+
+```sh
+mlx-bun serve <model> --temp 0 --top-p 0 --top-k 0 --max-tokens 512 --batch 1
+```
+
+- `--temp 0 --top-p 0 --top-k 0` — mlx_lm.server does not read
+  `generation_config.json` (we inject its sampling defaults, optiq-style);
+  its unset-request defaults are temperature 0.0, top-p/top-k off.
+- `--max-tokens 512` — its cap when a request omits `max_tokens` (ours is
+  none: run to EOS/stop or an explicit operator limit). The cap changes the stopping
+  point and `finish_reason`.
+- `--batch 1` — strict serial, arrival-independent numerics. (`--batch N`
+  is itself bit-parity with mlx-lm at B=N; pin 1 for golden regeneration.)
+- Already matching without flags: host/port, bf16 KV (`--kv-quant off`),
+  the L1 kernel preset (`--l1` is the default), logprobs caps, error text.
+
+These flags align sampling and caps. Exact generated trajectories also need
+matching prefill and cache boundaries. Stock MLX-LM HTTP splits system and
+thinking segments before reserving the final token. Different forward shapes
+can change near-tie greedy choices even inside the same Python model.
+The benchmark's explicit `--reference-prefill unsplit` control and the saved
+MiniCPM/Qwen investigation are documented in [benchmarks.md](benchmarks.md).
+
+The kill switches are bit-exact A/B levers; **the naked default is the L1
+set** (2026-07-05: an output-changing lever earns a default only by beating
+the L1 baseline in a paired A/B, and none did — the losing kernels were
+deleted, Phase 1 of
+[unified-engine-frontier-plan.md](../design/unified-engine-frontier-plan.md)).
+They are installed as `MLX_BUN_*` runtime values before the model loads,
+so they apply to `mlx-bun pi` too. They affect the **serial** decode path;
+the batch scheduler drives `model.forwardHidden` directly and is unaffected
+except by `--compiled-decode` at B=1 — see
 [Levers that don't reach the batched lane](#--batch-n-is-compat-mode--perf-flags-dont-apply-by-design).
 
+## `--isolate` semantics
+
+Design and measurements: [docs/reference/server-config.md](./server-config.md)
+(P1 single engine and P2 model pool landed). What the flag does, from
+`src/serve/isolate.ts` and the `serve` handler in `src/cli.ts`:
+
+- **Process layout.** The parent binds the TCP `--host`/`--port` and makes
+  zero MLX calls. It re-execs itself as `serve` with the same argv minus
+  the **parent-only flags** (`--isolate`, `--port`, `--host`, `--model-pool`,
+  `--no-open`), the model selectors replaced by a pinned `--model <path>`,
+  plus `--unix <socket>` (socket under the OS temp dir, named by parent
+  pid). Every other flag and every exported `MLX_BUN_*` variable reaches
+  the engine child unchanged (`env: process.env`); the child's stdout/stderr
+  are inherited, so load progress prints in your terminal.
+- **Readiness.** The parent polls the child's `/health` over the socket
+  (up to 15 minutes — large models take a while) and prints "engine ready
+  pid N". Requests arriving earlier wait on that readiness.
+- **Application state.** The web shell, job submissions, records, progress streams, settings and
+  Responses conversation history live in the CPU-only parent, so worker death or model eviction does not remove
+  them. Native inspection/merge/export and generation still execute in workers.
+  WebSocket chat retains its existing unsupported status under isolation.
+- **Proxying.** Remaining HTTP requests are forwarded over the unix
+  socket (hop-by-hop headers stripped, SSE streams through, a client
+  abort aborts the proxied fetch so the engine sees the disconnect).
+  `GET /engine` answers from the parent:
+  `{ isolated: true, pid, restarts, socket, response_store, pool?: { resident, default } }`.
+  Worker fields describe the currently resident default worker and are `null`
+  while it is evicted; inspection does not load a model. A recreated worker has
+  its own restart count.
+  `/ws/chat` is **not proxied** — `501` with a pointer to run without
+  `--isolate` for the web chat UI.
+- **Crashes.** A child exit (uncatchable Metal OOM/SIGTRAP included) is
+  respawned automatically (`restarts` increments), with at most three restarts
+  in a rolling 60-second window. Exhausting that budget leaves the host
+  unavailable until restarted. An engine that dies within 10 s of spawning
+  waits 5 s before the retry. In-flight requests
+  get `502 { error: { type: "engine_unavailable" } }`; bodyless `GET`/`HEAD`
+  requests are retried once after the respawn. A request whose client
+  disconnected answers `499`.
+- **Model pool (`--model-pool n`).** `POST` bodies on `/v1/chat/completions`,
+  `/v1/completions`, `/v1/messages`, `/v1/responses`, `/v1/embeddings` are
+  buffered to read `model`. An **exact** registry id (`/v1/models` id)
+  spawns/routes to that model's own child; anything else — empty, fuzzy,
+  `"gpt-4"`-style — rides the default engine (mlx-lm's ignored-field
+  semantics). Eviction over the cap: `POST /admin/drain` on the victim's
+  socket (gateway quiesce + demote its prompt cache to the SSD tier),
+  then stop; `/admin/drain` is unix-socket-only, never on the TCP
+  listener. Cold model starts are serialized through a bounded queue while the
+  resident model keeps serving; concurrent switches cannot evict a newly loaded
+  worker before its pending request is returned. `--model-pool` is clamped to ≥ 1.
+- **Managed GPU jobs.** A FIFO parent coordinator waits for active worker
+  responses, then holds a native execution lease inside every resident worker
+  before spawning a quantization/finetuning job. Startup, restart and eviction
+  participate in admission too. Job exit releases those leases. Dataset jobs use
+  the same loopback completion API and keep their progress in the parent.
+- **Shutdown.** `SIGINT`/`SIGTERM` cancels queued jobs, stops the active GPU job,
+  waits for in-process task cleanup, then closes workers and unlinks sockets.
+  Worker close also waits for pending startup admission and failed startup
+  attempts to release their execution leases, including before a child exists.
+
 ## Near-ceiling models on small machines (24 GB)
 
-A model whose weights approach the DEFAULT macOS GPU wired ceiling (~75% of
-RAM — e.g. a 17 GB model on a 24 GB machine) loads and serves short prompts,
-but a long prefill can exceed the ceiling and die with an **uncatchable**
-Metal OOM (the failure is asynchronous; no server error is possible).
-`mlx-bun serve` detects this at startup and prints the remedy:
+A model whose weights approach the default macOS GPU wired ceiling (~75%
+of RAM) loads and serves short prompts, but a long prefill can exceed the
+ceiling and die with an **uncatchable** Metal OOM. `mlx-bun serve` checks
+this at startup (model bytes > 80% of `maxRecommendedWorkingSetSize` while
+`iogpu.wired_limit_mb` is at its default) and prints the remedy:
 
 ```
 sudo sysctl iogpu.wired_limit_mb=<RAM_MB - ~2500>   # resets on reboot
 ```
 
 For very long contexts on such machines, additionally prefer `--kv-quant`
-(e.g. `turbo`) and expect the engine to use smaller prefill chunks — the
-prefill transient scales with chunk size × context depth.
+(e.g. `turbo`) — the prefill transient scales with chunk size × context.
 
 ## Per-request overrides
 
@@ -106,339 +351,496 @@ the server-wide default. Precedence, highest first:
    `min_p`, `xtc_probability`/`xtc_threshold`, `logit_bias`,
    `repetition_penalty`, `presence_penalty`/`frequency_penalty` (+ their
    `*_context_size` windows), `max_tokens`/`max_completion_tokens`, `stop`,
-   `chat_template_kwargs.enable_thinking`, `adapter`, `tools`)
-2. the matching `--temperature`/`--top-p`/`--top-k`/`--thinking` server default
+   `chat_template_kwargs.enable_thinking`, `reasoning_effort`, `adapter`,
+   `tools`)
+2. the matching `--temperature`/`--top-p`/`--top-k`/`--thinking`/`--max-tokens`
+   server default
 3. the model's `generation_config.json`
-4. the built-in fallback (temp 0.7, top-p/top-k 0 = off)
+4. the built-in fallback (temperature 0.7, top-p/top-k 0 = off)
 
 These ride all three protocol surfaces (`/v1/chat/completions`,
 `/v1/messages`, `/v1/responses`) because each funnels through one
-`handleChat`. Full field list: [server-api.md](./server-api.md).
+chat stage (`src/serve/chat-stage.ts`). Full field list: [server-api.md](./server-api.md).
+
+## Environment variables
 
-## Environment levers
+All `MLX_BUN_*` variables are captured once at process start into an
+immutable snapshot (`src/runtime-config.ts`); CLI flags install overrides
+into the same snapshot before the model loads. Boolean levers read the
+literal strings `"1"`/`"0"` — `flagOn` treats any other value as unset.
+Under `--isolate` the whole environment is inherited by the engine child.
 
-The CLI flags above set these; you can also export them directly (e.g.
-for paired A/B harnesses).
+### Serving levers (flag-backed and lane kill switches)
 
 | Env var | CLI flag | Default | Effect |
 | --- | --- | --- | --- |
-| `MLX_BUN_COMPILED_DECODE` | `--compiled-decode` | on (`!=="0"`) | Compiled decode graph replay. |
+| `MLX_BUN_COMPILED_DECODE` | `--compiled-decode` | on (`"0"` disables) | Compiled decode graph replay (serial lane; batch lane at B=1). |
+| `MLX_BUN_EARLY_FIRST_TOKEN` | — | off (`=1`) | Serial/native generation yields token zero before constructing the next decode step. This can reduce first visible output latency when token zero contains visible text. Later decode remains pipelined. Serial fill, grammar, checkpoint resume and single-token budgets retain their existing order. When a native consumer stops at the first yield and retains caller-owned caches, a non-aborted return completes that token’s M=1 forward before returning the cache. This preserves the ordinary pipeline’s boundary for later prefix reuse. Aborted requests do not start another forward; caches owned and disposed by the generation need no alignment. Native M1 packed-Qwen/MiniCPM/Gemma continuation gates pass; M4 packed/affine and serving acceptance remain. The continuous scheduler also yields after preparation creates its first active row when no other request is queued, allowing prepared output to flush before decode. It then rechecks cancellation, admission and shutdown; queued short admissions still group together. The setting is captured by generation and the batch runtime. Experimental pending broader cached/pressure and quiet-machine acceptance. |
+| `MLX_BUN_TURBOQUANT_FUSED_DECODE` | — | off (`=1`) | Experimental packed K/V decode fusion for an existing `--kv-quant turbo:...` cache, captured when each cache is created. A shared Metal operation unpacks keys and values, applies the existing key zero/scale and Lloyd-Max value scale, and preserves the codec's eager or deferred inverse rotation. It accepts supported bit widths, head dimensions 64/128/256/512 and 32/64-element groups with fp16/bf16/f32 metadata. Eager k8v3 with B1/H4, head dimension 256, fp16 metadata, group32 and at least 8192 cached tokens also fuses inverse rotation; other shapes keep the existing rotation path. Unsupported inputs, CPU streams and shapeless traces retain ordinary operations. Quantization, stored cache format and serving eligibility are unchanged. Joint-decoder Qwen native/serial and repeated long-context HTTP gates pass. The inverse operation passes both integrated Qwen model gates and six HTTP pairs per quant. MiniCPM/Gemma deferred-consumer serving gates also pass. The unreleased shared codec/layout supports ordinary and supported drafting groups without changing this kernel selection. Gemma now owns pre-write row positions across cache appends; the fix passes native checks on both Macs. Combined settings, pressure and strict M4 Pro acceptance remain. |
 | `MLX_BUN_NO_FUSED_SDPA` | `--fused-sdpa` (inverted) | follows `--kv-quant` | `=1` forces the stock unfused SDPA everywhere. |
-| `MLX_BUN_COMPILED_GEGLU` | `--compiled-activations` | on (`!=="0"`) | Gemma geglu via mlx-lm's `@mx.compile` closure — the faithful default (bit-exact vs mlx-lm, one kernel). `=0` → uncompiled composition (same parity, slower). |
-| `MLX_BUN_COMPILED_SWIGLU` | `--compiled-activations` | on (`!=="0"`) | `mx.compile`'d SwiGLU (`silu(gate)·up` → one kernel) on MiniCPM5 decode (M=1), porting mlx-lm's `activations.py`. Bit-exact (passes the exact logit-parity gate), both lanes. +5.5% CPM5 decode. (qwen3/qwen3.5/universal compile swiglu unconditionally, independent of this flag.) |
-| `MLX_BUN_FORCE_WIRE` | `--force-wire` | off (`==="1"`) | Wire weights for the generation. |
-| `MLX_BUN_PAGED_KV` | `--paged-kv` | off (`==="1"`) | Paged KV cache (see the flag row above — v1: serial, Gemma4, bf16; refusals + prompt-cache bypass apply identically). |
-| `MLX_BUN_ALLOW_PRIVATE_MEDIA` | `--allow-private-media` | off (`==="1"`) | Permit `image_url`/`audio_url` fetches to private/loopback/link-local hosts (the SSRF guard's LAN escape hatch — see the flag row above; timeout + size cap still apply). |
-| `MLX_BUN_PREFILL_TAIL_SPLIT` | — | on (`!=="0"`) | Oracle prefill convention (mlx-lm `generate_step` and its server's batched engine): drain the prompt only to len−1, then compute step-0 logits from a separate **L=1 forward of the last prompt token**. Both lanes. **The SPEC lane (`--draft-model`) follows its own oracle's shape under the same flag** (mlx-lm `speculative_generate_step`, re-anchored 2026-07-07): both target AND draft drain to len−1 and there is **no separate step-0 at all** — the un-drained last prompt token heads the first verify window. Gated live: token-for-token vs the oracle venv incl. a knife-edge cell (tests/spec-serve.test.ts "L1 knife-edge"). (The standalone optiq-oracled assistant loop `specGenerate` deliberately keeps FULL-prompt prefill — optiq's own convention.) `=0` restores the pre-2026-07-07 full-final-chunk convention everywhere (A/B lever + kill switch) — that convention is ulp-different in bf16 at step 0 AND in the last prompt token's stored KV, which flips near-tie greedy streams vs mlx-lm (the 2026-07-07 12B completion-probe divergence). |
-| `MLX_BUN_GRAMMAR` | — | on (`!=="0"`) | `=0` disables structured-output mask compilation. Requested constraints still take the documented graceful-degrade route: chat prompt injection plus a `Warning` header; raw completions emit the header only. |
-| `MLX_BUN_GRAMMAR_JUMP` | — | **off** (`==="1"`) | Jump-forward decoding for structured output (SGLang's technique via xgrammar's `findJumpForwardString`): when the grammar forces a unique continuation (JSON punctuation/keys; longest under `any_whitespace:false` fixed formatting), the SERIAL lane emits its retokenized ids with ONE multi-token forward instead of one masked forward per token. Lossless in STRING space and always grammar-valid, but the token stream (and, conditioning on it, content after a forced span) can legally differ from an unjumped run — no oracle, hence opt-in. Excluded automatically when either `logprobs` or `top_logprobs` is requested, because jumped tokens have no sampled distribution row; SentencePiece-family tokenizers whose raw encode can't reproduce a mid-stream span simply never jump (guard degrades to normal decode). The batch lane's per-row grammar (`#stepGrammar`) does not jump yet. Tests: tests/grammar-jump.test.ts (contract + Llama-3.2-1B e2e). |
-| `MLX_BUN_SSD_WRITEBEHIND` | — | on (`!=="0"`) | `--ssd-cache`'s debounced write-behind snapshot (restart survival). `=0` disables it entirely — the paired-A/B lever + kill switch for the 2026-07-07 idle-gating fix (eviction/demotion spills still write). The flush is **idle-gated**: every per-tensor step (blocking GPU sync + `writeSync`) waits for the engine to go idle and pauses when a request arrives mid-flush, so durability work never taxes an active decode (pre-fix: a ~16k entry's flush overlapping cached ctx repeats cost e4b ~9% decode@ctx vs mlx-lm). The snapshot timer also never holds batch admission — while rows are active it re-arms instead of grabbing the exclusive. |
-| `MLX_BUN_SSD_SPILL_QUEUE_GB` | — | 2 | Byte cap on the write-behind queue's PENDING spill clones (they pin the evicted entries' GPU memory until the idle-gated flush runs). Over cap, the oldest queued spill drops and its clones free immediately. The durability coordinator keeps the source snapshot dirty so an explicit or shutdown flush can retry it. `=0` keeps only the newest + in-flight clone pinned (the soft cap never drops the item just enqueued). `/stats.ssd_cache` reports pending, dropped, failed, and longest-durable-prefix counters. |
-| `MLX_BUN_SHUTDOWN_TIMEOUT_MS` | — | `120000` | Maximum time `mlx-bun serve` gives active requests plus the SSD durability flush after `SIGINT` or `SIGTERM`. On timeout the process logs the remaining snapshot/spill counters and exits. |
-| `MLX_BUN_BATCH_NO_PIPELINE` | — | off (`==="1"`) | Disables the batch lane's pipelined token readback (the A/B lever for the shipped pipelining, like its peers `MLX_BUN_BATCH_EXTEND` / `MLX_BUN_BATCH_SSM` / `MLX_BUN_BATCH_VEC_SAMPLE`). Numerically equivalent, slower steps. |
-| `MLX_BUN_DSPARK_MINCONF` | — | checkpoint-driven | Overrides the DSpark/DeepSpec draft scheduler's minimum-confidence threshold (Alg-1 draft-length pruning). Spec lane only; no CLI flag — a calibration/A/B lever (see [dspark-speculative-decoding.md](../design/dspark-speculative-decoding.md)). |
-| `MLX_BUN_EXPERT_OFFLOAD` | `--expert-offload` | off (`==="1"`) | Env mirror of the flag for direct/library runs (scripts that never parse serve flags). |
-
-(`MLX_BUN_PERF_KERNEL`, `MLX_BUN_FUSED_GELU`, `MLX_BUN_FUSED_DECODE`,
-`MLX_BUN_FUSED_SWIGLU*`, and `MLX_BUN_CPM5_FAITHFUL` were deleted
-2026-07-05 with their kernels — see
-[unified-engine-frontier-plan.md](../design/unified-engine-frontier-plan.md);
-exporting them now does nothing.)
+| `MLX_BUN_COMPILED_GEGLU` | `--compiled-activations` | on (`"0"` disables) | Gemma geglu via mlx-lm's `@mx.compile` closure. `=0` → uncompiled composition (same parity, slower). |
+| `MLX_BUN_COMPILED_SWIGLU` | `--compiled-activations` | on (`!== "0"`) | Compiled SwiGLU on MiniCPM5 decode and prefill, matching mlx-lm. An enclosing compiled graph owns its own fusion. qwen3/qwen3.5/universal compile unconditionally. |
+| `MLX_BUN_FORCE_WIRE` | `--force-wire` | off (`=1`) | Wire weights for the generation. |
+| `MLX_BUN_TRELLIS` | — | `kernel` (`=expand`) | Packed trellis-coded weights (`mode: "trellis"` modules, Q2b — design: `docs/design/turboquant.md`). `kernel` serves them through the Metal decode kernels (M≤4 matvec; larger M expands one tensor to bf16 and runs a stock matmul). `=expand` decodes every trellis tensor at LOAD into 8-bit g64 affine (+~4 GiB at 27B, the eval-carrier numerics) and serves it through the stock quantized path — the fallback for a machine where the kernels lose. |
+| `MLX_BUN_TRELLIS_VARIANT` | — | `13` | Trellis kernel variant (see [Q2b experiments](../design/turboquant.md)): `6` = code computed inline × reciprocal, weight served as f32 code×scale; `1` adds a residual step and bf16 rounding to reproduce the fake-quant artifact's weights. `0`/`2`/`3`/`4`/`5` are bench-only decoder variants. Variants `7`–`13` retain variant-6 decode values; `13` is the measured default. These variants tune work assignment; `11` tiles eligible short axis-1 prefills, `12` adds split-K axis-0 prefill at M=5..8, and `13` also vectorizes remaining bf16 expansion for k2/k3/k4, T=256, L=12. On MLX 0.32.2/M3+ the axis-1 M5..15 path instead uses direct packed decoding with native wide-matvec arithmetic when Qwen's RMSNorm establishes aligned row-contiguous inputs; callers without that layout proof use native expansion/matmul. Qwen27B additionally selects an integer codebook for its interleaved k3 down projection at M=3/4 with bf16 activations; other calls retain the computed decoder. Prefill rounds weights to the activation dtype. |
+| `MLX_BUN_TRELLIS_ASYNC_EXPAND` | — | off (`=1`) | Experimental variant-13 expansion scheduling. Submit an expanded projection asynchronously while MLX active allocation is below 75% of the device's recommended working set; retain blocking evaluation above it. Other variants are unchanged. Uses the execution's runtime-policy snapshot and preserves the caller's layer barriers. The threshold is not a total-memory cap. Qwen27B integrated native, repeated serial/continuous HTTP and both saved-agent pressure gates pass on M4 Pro 24 GB. Broader-model, combined-optimization and quiet M4 Pro acceptance remain. |
+| `MLX_BUN_PAGED_KV` | `--paged-kv` | off (`=1`) | Paged KV cache; the same refusals and prompt-cache bypass as the flag. |
+| `MLX_BUN_ALLOW_PRIVATE_MEDIA` | `--allow-private-media` | off (`=1`) | Permit media fetches to private/loopback/link-local hosts (timeout + size cap still apply). |
+| `MLX_BUN_EXPERT_OFFLOAD` | `--expert-offload` | off | `=<dir>` — the path of a built expert-offload file, activated at module load (`src/expert-offload.ts`) for scripts and library runs that never parse serve flags. The CLI flag builds the file and activates it itself. |
+| `MLX_BUN_PREFILL_TAIL_SPLIT` | — | on (`"0"` disables) | Oracle prefill convention: drain the prompt to len−1, then compute step-0 logits from a separate L=1 forward of the last prompt token (mlx-lm `generate_step` and its batched engine). Both lanes. The spec lane follows its own oracle's shape under the same flag (mlx-lm `speculative_generate_step`: target and draft drain to len−1, no separate step 0). `=0` restores the full-final-chunk convention everywhere — ulp-different at step 0, flips near-tie greedy streams vs mlx-lm. |
+| `MLX_BUN_GRAMMAR` | — | on (`"0"` disables) | `=0` disables structured-output mask compilation; requested constraints take the graceful-degrade route (chat prompt injection + a `Warning` header; raw completions header only). |
+| `MLX_BUN_TOKEN_MASK` | `host` \| `metal` | `host` | Experimental grammar-mask implementation. `metal` uploads packed bits and applies the same additive mask in a compiled shared Metal kernel. Captured when the grammar controller is created; serving performance gates remain open. |
+| `MLX_BUN_GRAMMAR_JUMP` | — | **off** (`=1`) | Jump-forward decoding for structured output (SGLang's technique via xgrammar's `findJumpForwardString`): when the grammar forces a unique continuation, the **serial** lane emits its retokenized ids with one multi-token forward. Lossless in string space and always grammar-valid, but the token stream can legally differ from an unjumped run — no oracle, hence opt-in. Skipped when `logprobs`/`top_logprobs` is requested; SentencePiece-family tokenizers that can't reproduce a mid-stream span never jump. The batch lane does not jump. |
+| `MLX_BUN_FILL` | — | **off** (`=strict` \| `=echo`) | **Token fast-forwarding** — lookup, not speculation. Spans of the assistant turn that the request's `tools` + the chat template already DETERMINE (tool-call open scaffold, the rest of a tool name after its first disambiguating token, a sole-required-key skeleton) are appended as committed context in bounded model forwards, and sampling resumes after them. No draft, no verify, no rollback — an injected token is context, indistinguishable to the model from one it sampled. Rows are compiled per request by diffing probe renderings of the model's OWN template, always sliced from a rendering carrying the REAL tool names and schema keys (so a span can never split a merged token such as Qwen3.5's `=get`), so a template that does not render `tool_calls` yields no rows and no fill. Key and call-close assertions require a closed object schema with exactly one required property and no pattern properties; optional/additional arguments stay available. If no property is required, the compiler preserves the empty-argument choice. Template-derived rows require a fresh request-local parser context. Tool-only output after closed reasoning qualifies; quoted/fenced examples, incomplete delimiters and ambiguous literals decline. Requests with more than 32 tools decline template-derived fills. Held-out identity coverage remains open. **Serial lane only** (it never forces a request off the batch lane) and refused for: fixed `seed`, `logprobs`/`top_logprobs`, structured output (grammar owns forced tokens), image/audio/video prompts, a mounted draft model, quantized/TurboQuant KV, and sliding-window models (one warning). Injection bypasses the sampler — a behavior-policy deviation at `temperature > 0` — hence opt-in. `=echo` additionally arms the **echo index** (Lab tier): spans copied from earlier in the same session, held under policy `verify` — the engine appends them in the same single forward, reads the argmax already in that forward's logits at every span position (free, no extra pass), keeps the prefix the model agrees with, and rewinds the rest through the same cache contract the spec lane's rounds use. A wrong echo costs a rewound forward, never a wrong token. Strict rows stay policy `assert` (no readback, no checkpoint). Telemetry: `usage.fill`. |
+| `MLX_BUN_FILL_MAX_SPAN` | — | `32` (floor 2) | Hard cap on one injected span. |
+| `MLX_BUN_FILL_APPEND_CHUNK_SIZE` | — | `0` (model limit) | Execution chunk cap for assert-policy fill, captured by each FillSession. Zero selects the model's qualified limit; a positive value can lower that limit. The Qwen 27B text path on `applegpu_g16s`, validated on the M4 Pro, supports up to four positions with supported affine projections and optional Trellis MLPs. It rechecks the limit after each chunk and splits at MLX's attention arithmetic boundaries to preserve one-token numerics. Other configurations use one position per forward. The engine commits the complete span before emission and resumes sampling after it, with no intermediate vocabulary heads or verification. Verify-policy proposals keep a single forward for recurrent rollback. The injected-token cap is unchanged. Library callers can set `FillSession`'s `appendChunkSize` option. Held-out, broader-model and quiet M4 Pro acceptance remain. |
+| `MLX_BUN_FILL_K` | — | `8` (floor 2) | Echo anchor length: the k-gram that must match before a copied span is proposed (the corpus study's token-level threshold). |
+| `MLX_BUN_FILL_CANDIDATES` | — | `24` | Echo bucket-scan cap — the NEAREST occurrences of the anchor win. A span extends only while every one of them agrees; the first disagreement (a **branch point**) ends it. |
+| `MLX_BUN_FILL_INDEX_MAX` | — | `131072` (floor 1024) | Token cap on the growing echo index; past it the index freezes rather than growing. |
+| `MLX_BUN_FILL_TRACE` | — | off (`=1` \| `=<file.jsonl>`) | `=1`: assert the fill cache-alignment invariant (`cache offset == prompt + forwarded`) on both sides of every append, and log each event. `=<file.jsonl>`: additionally append one record per proposal — proposed ids/text next to what the model's own logits said at every span position (position 0 = the in-flight sample, then argmax), for both policies; under `assert` that readback is trace-only. `bun scripts/fill.ts trace <file>` prints the list and per-position agreement. |
+| `MLX_BUN_GRAMMAR_BATCH` | — | on (`"0"` forces serial) | `=0` routes grammar requests to the serial lane instead of per-row batch matchers (A/B + kill switch). |
+| `MLX_BUN_BATCH_SSM` | — | on (`"0"` forces serial) | `=0` excludes SSMCache (Qwen3.5 gated-DeltaNet hybrids) from the batch capability gate → those models route serial. |
+| `MLX_BUN_BATCH_EXTEND` | — | on (`"0"` reverts) | Joining rows append to the running batch's KV in one pad+concat (mlx-lm `BatchKVCache.extend`). `=0` reverts to whole-batch re-merge (numerically equivalent, O(B·S)). |
+| `MLX_BUN_BATCH_VEC_SAMPLE` | — | on (`"0"` reverts) | Vectorized greedy batch sampling; `=0` falls back to per-row sampling (bit-equal A/B). |
+| `MLX_BUN_BATCH_NO_PIPELINE` | — | off (`=1`) | Read each batch step's tokens synchronously instead of pipelined (A/B lever; numerically equivalent, slower). Read once at module load. |
+| `MLX_BUN_SSD_WRITEBEHIND` | — | on (`"0"` disables) | `--ssd-cache`'s debounced write-behind snapshot (restart survival). Each tensor write waits for an idle engine, then owns the generation lock for that step. Pending writes do not drain active batches or delay request admission for a full flush. `=0` disables it; eviction and demotion spills still write. |
+| `MLX_BUN_SSD_SPILL_QUEUE_GB` | — | `2` (GiB; any finite value ≥ 0) | Byte cap on pending write-behind spill clones (they pin evicted entries' GPU memory until the generation-locked flush runs). Over cap the oldest queued spill drops and frees immediately; the durability coordinator keeps its snapshot dirty so an explicit or shutdown flush can retry. `=0` keeps only the newest + in-flight clone pinned. Counters in `/stats.ssd_cache`. |
+| `MLX_BUN_SHUTDOWN_TIMEOUT_MS` | — | `120000` (any finite value > 0) | Maximum time `serve` gives active requests plus the SSD durability flush after `SIGINT`/`SIGTERM`; on timeout it logs the remaining snapshot/spill counters and exits. |
+| `MLX_BUN_DSPARK_MINCONF` | — | checkpoint-driven | Overrides the DSpark/DeepSpec draft scheduler's minimum-confidence threshold (draft-length pruning). Spec lane only ([docs/design/speculative-decoding.md](../design/speculative-decoding.md)). |
+| `MLX_BUN_MEMORY_BATCH` | — | `1` (integer ≥ 1) | Row width for `mlx-bun memory` model calls. 1 = serial (batching measured 1.7–1.9× slower on the heterogeneous extract/chunk workload — [docs/design/dreaming-nightly-pipeline.md](../design/dreaming-nightly-pipeline.md)). |
+
+### Paths and tooling
+
+| Env var | Default | Effect |
+| --- | --- | --- |
+| `MLX_BUN_LIBMLXC` | resolved: beside the executable → native-pack cache (`~/Library/Caches/mlx-bun/native-v<ver>-<arch>/`) → Homebrew | Explicit path to `libmlxc.dylib`; its directory is then the native runtime dir (`src/mlx/ffi.ts`, `src/native-pack.ts`). |
+| `MLX_BUN_NATIVE_PACK_URL` | the GitHub release asset for `NATIVE_PACK_VERSION` | Override the native runtime pack download URL (first-run fetch, sha256-verified). |
+| `MLX_BUN_FRAME_EXTRACT` | resolved: beside the executable → native-pack cache → `dist-native/` → compile from `src/native` | Explicit path to the `mlx-bun-frame-extract` AVFoundation sidecar used for video input. |
+| `MLX_BUN_EXPERT_IO_DYLIB` | resolved: beside the executable → native-pack cache → `dist-native/` | Explicit path to `libmlx_bun_expert_io.dylib` (GLM-5.2 streamed expert I/O). |
+| `MLX_BUN_WIKI` | `~/.mlx-bun/wiki` | Root of the memory vault (`~` expands). |
+| `MLX_BUN_JOBS_DB` / `MLX_BUN_JOBS_DIR` | the `JobStore` defaults | Job-store SQLite path and logs dir. The job runner sets both on the subprocess it spawns so the child opens the same store; override for tests. |
+| `MLX_BUN_CALIBRATION_JSONL` | bundled optiq mix in the oracle venv | Explicit calibration corpus for `mlx-bun convert`/quantize. |
+| `MLX_BUN_ORACLE_PYTHON` | `/Users/joshrossi/Code/mlx-lm/.venv/bin/python` (the reference box; machine-specific, never re-commit) | Oracle venv interpreter for evals that shell out to Python (HumanEval). |
+| `MLX_BUN_EVAL_DATA` | `~/.cache/mlx-bun/eval-data` | Directory of exported eval `.jsonl` datasets. |
+
+### Diagnostics and tracing (off unless set)
+
+| Env var | Value | Effect |
+| --- | --- | --- |
+| `MLX_BUN_P2R_TRACE` | `=1` | Per-request prompt→response phase trace (admission wait, prefill, token-zero, …) for `/v1/chat/completions` and `/v1/completions`; records print to stderr as JSON lines. The trace id is `x-mlx-bun-trace-id` when the request sends it. |
+| `MLX_BUN_P2R_SYNC` | `=1` (with `MLX_BUN_P2R_TRACE`) | Attribution mode: synchronizes the GPU at phase boundaries so each phase is charged its own work (MLX is lazy — without it, unsubmitted work lands in the next phase). Slows the traced request. |
+| `MLX_BUN_LANE_DEBUG` | `=1` | Logs each request's scheduling placement (`mechanism` + shape) to stderr. |
+| `MLX_BUN_BATCH_STEP_TRACE` | `=1` | Per-step phase timing in the batch scheduler (build / read / emit / gap), read once at module load; summarized by `stepTraceReport()`. |
+| `MLX_BUN_GRAMMAR_DEBUG` | `=1` | Logs per-step grammar row state in the batch scheduler. |
+| `MLX_BUN_PREFILL_MEM_LOG` | `=1` | Logs active/peak memory after each serial prefill chunk. |
+| `MLX_BUN_EXPERT_TRACE` | `=<path>` | Records every MoE router decision as JSONL to that path (adds a per-call GPU→host sync — a measurement tool, not a serving path). |
+| `MLX_BUN_PI_DEBUG` | any non-empty | Extra `[pi-web]` logging (prompt fingerprint, tool/memory surface) for the web-chat pi session. |
+| `MLX_BUN_EVAL_DEBUG` | `=1` | HumanEval: pipe the sandbox's stderr and print failures. |
+
+### Eval and training
+
+| Env var | Default | Effect |
+| --- | --- | --- |
+| `MLX_BUN_EVAL_THINK` | off (`=1`) | Evaluate with `enable_thinking` on (default off for parity with optiq's published non-thinking numbers). |
+| `MLX_BUN_EVAL_KV_QUANT` | off (`=1`) | Generate through the model's `kv_config.json` quantized KV during evals (default full-precision KV, matching optiq's eval). An explicit `--kv-quant` on `mlx-bun eval` wins. |
+| `MLX_BUN_MMLU_FROZEN` / `MLX_BUN_GSM8K_FROZEN` / `MLX_BUN_HUMANEVAL_FROZEN` / `MLX_BUN_IFEVAL_FROZEN` / `MLX_BUN_BFCL_FROZEN` / `MLX_BUN_HASHHOP_FROZEN` | on (`"0"` reverts) | Use the frozen oracle-exported problem sets (parity with the reference numbers); `=0` falls back to our own sampling of the full set. |
+| `MLX_BUN_TRAIN_ATTN` | `ops.sdpa` (unset) | `=flash` selects the hand-rolled O(L)-memory flash kernel for training attention (~30× slower). **Refused for Gemma models** (e4b SIGTRAPs at seq ≥ 2048); MiniCPM5 allowed. |
+| `MLX_BUN_FLASH_MIN_M` | `1024` | With a flash-CCE head requested, rows shorter than this take the exact fused head instead; `=0` always honors flash. |
+| `MLX_BUN_MEM_LOG` | off (`=1`) | Trainer per-step memory logging. |
+
+`MLX_BUN_CCE_*` (flash-CCE kernel variant selectors and epsilons in
+`src/train/flash-cce.ts`) and `MLX_BUN_SEG_*` (segmented-backward head and
+memory probes in `src/train/segmented.ts`) are intentionally undocumented
+developer knobs; `MLX_BUN_EVAL_KV_ARM_NOOP` is an inert placeholder the KL
+harness flips.
+
+### Removed
+
+`MLX_BUN_PERF_KERNEL`, `MLX_BUN_FUSED_GELU`, `MLX_BUN_FUSED_DECODE`,
+`MLX_BUN_FUSED_SWIGLU*`, `MLX_BUN_CPM5_FAITHFUL`, and the `--l3` tier were
+deleted 2026-07-05 with their kernels
+([unified-engine-frontier-plan.md](../design/unified-engine-frontier-plan.md));
+exporting the variables now does nothing, and `--l3` exits with a pointer.
 
 ## Execution modes: serial vs. `--batch N`
 
-Scheduling is **concurrency-driven** (default cap 8, flipped 2026-07-05): a
-lone scheduler request uses the B=1 fast path — its caches are adopted
-serial-class objects, compiled decode replays, and the prompt cache and SSD
-tier serve it. Only a second concurrent request causes a batch layout to
-exist. The flag declares the concurrency cap; active rows select B=1 or B=N
-inside the scheduler.
+Scheduling is **concurrency-driven** (default cap 8): a lone scheduler
+request uses the B=1 fast path — its caches are adopted serial-class
+objects, compiled decode replays, and the prompt cache and SSD tier serve
+it. Only a second concurrent request causes a batch layout to exist. The
+flag declares the concurrency cap; active rows select B=1 or B=N inside
+the scheduler.
+
+No flag is needed to enable this default. The cap is a maximum, not a minimum
+group size: a lone eligible request starts at B=1. The serial executor remains
+available through `--batch 1` and for unsupported request compositions.
 
 `--batch 1` pins the strict serialized single-queue path: one generation
 at a time, arrival-independent numerics (a request's bits never depend on
 what else was in flight). Pin it for golden regeneration and
-reproducibility work. Note the trade is real but small: batched rows
-carry bf16 left-pad reduction-order noise vs their solo runs (calibrated
-per model in the gated suites) — a request that never shares a step is
-bit-identical either way.
+reproducibility work. Batched rows carry bf16 left-pad reduction-order
+noise vs their solo runs (calibrated per model in the gated suites) — a
+request that never shares a step is bit-identical either way.
 
 The batched engine is **bit-parity with `mlx_lm.server` at `B=N`** per
-row for bf16, and per-row oracle-gated for the quantized compositions
-(Phase 3.1 + milestone 2 — compositions mlx-lm does not ship):
+row for bf16, and per-row oracle-gated for the quantized compositions:
 
-- **KV quant unset ⇒ bf16** — the batch path engages out of the box
-  (bf16 is the serial default too, since 2026-07-05).
+- **KV quant unset ⇒ bf16** — the batch path engages out of the box.
 - **`--kv-quant config`** (per-layer mixed precision) **batches** on
-  every shipped model — full-attention layers since Phase 3.1 (cpm5),
-  rotating layers since milestone 2 (gemma) — applied per row, gated
-  bit-exact/KL-0 for unpadded rows vs the serial composition.
-- **`--kv-quant 4|8`** (uniform bits) routes those requests to the
-  serial lane (quantizedKvStart threshold semantics; a startup warning
-  is printed).
+  every shipped model — full-attention and rotating layers — applied per
+  row, gated bit-exact for unpadded rows vs the serial composition.
+- **`--kv-quant 4|8`** uses shared affine conversion and batches on supported
+  full-attention, rotating and hybrid recurrent models.
+- **`turbo`** uses encoded shared rows for qualified start-zero full-attention caches; sliding layers stay bf16.
 
 ### Scheduling declaration (`GenerationGateway.place`)
 
-`--batch` declares a concurrency cap. With a cap greater than one,
-`GenerationGateway.place()` freezes one scheduling mechanism for the exact
-resolved execution composition. `continuous` admits it to the scheduler; that
-scheduler chooses its B=1 fast path or B=N step from active rows. `serial`
-preserves the strict or dedicated executor. The support check never removes,
-downgrades, or substitutes MTP, KV schemes, TurboQuant, grammar, adapters, or
-sampling. A composition the scheduler does not implement drains it and runs
-unchanged through the serial mechanism.
-
-The two mechanisms are mutually exclusive on the GPU (one `AsyncMutex`), so a
-serial request never runs alongside a continuous step, while requests admitted
-to the continuous scheduler can run concurrently with each other.
+With a cap greater than one, `GenerationGateway.place()` freezes one
+scheduling mechanism for the exact resolved request shape: `continuous`
+admits it to the scheduler (which then picks B=1 or B=N from active rows);
+`serial` preserves the strict executor. The support check never removes,
+downgrades, or substitutes MTP, KV schemes, TurboQuant, grammar, adapters,
+or sampling — a composition the scheduler does not implement drains it and
+runs unchanged serially. The two mechanisms are mutually exclusive on the
+GPU (one `AsyncMutex`).
 
 | Request property | Continuous scheduler? |
 | --- | --- |
-| vision (image parts) | ❌ serial — needs offset-0 single-seq prefill + bidirectional image mask |
-| LoRA `adapter` (resolves to ≥1) | ❌ serial — `loraState.active` is one per-generation field; per-row adapters unsupported |
-| `logprobs` / `top_logprobs` | ❌ serial — the batched sampler doesn't capture logprob arrays yet |
-| explicit `seed` | ❌ serial — reproducibility ⇒ solo (matches mlx-lm) |
-| KV quant active (explicit `--kv-quant`) | ✅ batches for per-layer `config` schemes — full-attention layers (Phase 3.1) AND rotating layers (milestone 2: gemma's kv_config; unpadded row gated KL-0 vs serial); uniform bits → serial |
-| `--kv-quant turbo` (TurboQuant) | ❌ serial, unconditionally — solo-only in v1 (novel `Cache` class, no batched merge/filter/temporalView; belt-and-braces refusal in placement on top of the automatic capability-gate exclusion) |
-| `--draft-model` mounted | ❌ serial, server-wide — speculation is a B=1 latency mode (upstream `is_batchable = draft is None`) |
-| `repetition_penalty` / `min_p` / `xtc_*` / `logit_bias` / presence+frequency penalties | ✅ batches — per-row logits processors over a per-row device-side history (since 2026-07-02; some models — Qwen3.5 — ship a *default* repetition penalty, which used to route everything serial) |
-| structured output (`response_format` / `guided_*`) | ✅ batches — per-row grammar matchers driven by the scheduler (`MLX_BUN_GRAMMAR_BATCH=0` forces serial) |
+| vision / audio / video parts | ❌ serial — offset-0 single-sequence prefill + media masks |
+| LoRA `adapter` (resolves to ≥1) | ✅ identical ordered adapter sets share a group; different sets wait for the next group |
+| `logprobs` / `top_logprobs` | ✅ ordinary and qualified MTP/lookup/standalone-draft groups capture per-request probabilities; other exclusions still apply |
+| explicit `seed` | ✅ request-local random stream; reproducibility also depends on model arithmetic and batch composition |
+| KV quant active | ✅ batches for per-layer `config` and uniform `4`/`8` schemes on supported cache layouts |
+| `--kv-quant turbo` | ✅ qualified TQ layouts; ordinary and supported grouped drafting library requests also support delayed row-local conversion |
+| `--draft-model` mounted | ✅ qualified MTP, lookup, standalone, assistant, DeepSpec and DSpark/DFlash groups; provider and request capabilities determine eligibility; supported ordinary requests batch |
+| `repetition_penalty` / `min_p` / `xtc_*` / `logit_bias` / presence+frequency penalties | ✅ batches — per-row logits processors over per-row device-side history (Qwen3.5 ships a *default* repetition penalty, so this is load-bearing) |
+| structured output (`response_format` / `guided_*`) | ✅ batches — per-row grammar matchers (`MLX_BUN_GRAMMAR_BATCH=0` forces serial) |
 | `temperature` / `top_p` / `top_k` | ✅ batches (each row samples with its own seed) |
-| `stop` sequences | ✅ batches (per-row `StopMatcher` in the onToken closure) |
-| `tools` / `tool_choice` | ✅ batches (per-row tool router; decode-layer parse) |
-| `--thinking` / `enable_thinking` | ✅ batches (template-render concern, lane-independent) |
-| multi-turn / long prompt | ✅ batches **with prompt-cache reuse** (Phase 3.2): a joiner restores the longest usable cached prefix and prefills only the suffix (`cached_tokens` reported); a request that finishes without ever sharing the batch puts its caches back. **Prefix sharing (2026-07-05): serves are non-consuming clones**, so concurrent agents with a shared system prompt all reuse one prefill and nobody's entry is destroyed. A row that merges with others is not re-put (its entry ages out) — heavy concurrency reduces hit rate, never correctness |
-
-**Which models use continuous scheduling:** full-attention (CPM), sliding-window (Gemma),
-hybrid gated-DeltaNet (Qwen3.5 — the SSM batched path, token-exact vs the
-mlx-lm B=2 oracle, landed 2026-07-02; `MLX_BUN_BATCH_SSM=0` reverts it to
-serial routing), and **plain full-attention Tier-0 universal archs**
-(Llama etc. — per-row RoPE gated token-exact vs mlx-lm B=2, 2026-07-03).
-Still serial by the model-level capability gate: gemma2-family and
-sliding-window *universal* archs (unvalidated cells) and DiffusionGemma
-(non-autoregressive).
-
-A request requiring the serial mechanism **drains** the scheduler: while it waits, the
-scheduler stops admitting new rows, finishes the running ones, and
-releases the GPU so the serial request runs (mlx-lm's `drain_batch`);
-admission then resumes. So a steady stream of batchable traffic cannot
-starve a serial-lane request.
-
-Placement is not a load-dependent fallback. A request admitted to the
-continuous scheduler stays there whether it is the only active row or one of
-many; arrivals select B=1 versus B=N *inside* that mechanism. `--batch 1`
-remains the arrival-independent numerics pin. This matches mlx-lm's
-concurrency-driven serving contract while keeping the strict serial route for
-goldens and compositions that need it.
+| `stop` sequences | ✅ batches (per-row `StopMatcher`) |
+| `tools` / `tool_choice` | ✅ batches (per-row tool router) |
+| `--thinking` / `enable_thinking` | ✅ batches (template-render concern) |
+| multi-turn / long prompt | ✅ batches with prompt-cache reuse: a joiner restores the longest usable cached prefix and prefills only the suffix (`cached_tokens` reported) |
+
+**Which models use continuous scheduling** (`#modelCachesBatchable`):
+full-attention (CPM), sliding-window (Gemma), hybrid gated-DeltaNet
+(Qwen3.5 — `MLX_BUN_BATCH_SSM=0` reverts to serial routing), and plain
+full-attention Tier-0 universal archs (Llama etc., gated token-exact vs
+mlx-lm B=2). Still serial by the capability gate: gemma2-family
+(`maskArray`) and sliding-window *universal* archs (unvalidated cells) and
+DiffusionGemma (non-autoregressive).
+
+A request requiring the serial mechanism **drains** the scheduler: while
+it waits, the scheduler stops admitting new rows, finishes the running
+ones, and releases the GPU (mlx-lm's `drain_batch`); admission then
+resumes, so batchable traffic cannot starve a serial-lane request.
+Placement is not load-dependent: a request admitted to the continuous
+scheduler stays there whether it is the only row or one of many.
 
 ## Compatibility matrix
 
-How each option behaves under each configured mechanism. "serial" =
-`--batch 1`; `--batch N` permits continuous scheduling, although a declared
-composition may still require the serial mechanism as shown above.
+"serial" = `--batch 1`; `--batch N` permits continuous scheduling, though a
+declared composition may still require the serial mechanism as shown above.
 
 | Option | serial (`--batch 1`) | `--batch N` (N>1) |
 | --- | --- | --- |
 | `--kv-quant config` | ✅ applied to all requests | ✅ batches where the loaded cache capability supports the per-layer scheme; otherwise routes serial |
-| `--kv-quant 4`/`8` | ✅ applied to all requests | ⚠️ uniform-threshold semantics route requests serial |
-| `--kv-quant turbo[:k<bits>v<bits>]` | ✅ applied to all requests | ⚠️ applied, but forces **all** requests to the serial lane (no batching) |
-| `--kv-quant off` | ✅ bf16 | ✅ bf16 (same as the implicit batch default) |
-| *(kv-quant unset)* | bf16 (the L1 default) | **bf16** (Option B) — incl. serial-lane fallback requests |
-| `--memory-budget` | ✅ per-request admission | ✅ per-request admission — but **not aggregate** across rows (see limitations) |
-| `--prompt-cache` | ✅ prefix reuse | ✅ prefix reuse on BOTH lanes (Phase 3.2): batch joiners `take()` at admission; lone never-merged rows `put()` back on finish (merged rows' entries age out) |
+| `--kv-quant 4`/`8` | ✅ applied to all requests | ✅ server start=0 converts during prefill and batches on supported cache layouts; delayed library thresholds still use serial |
+| `--kv-quant turbo[:k<bits>v<bits>]` | ✅ ordinary decode; strict serial speculation remains excluded | ✅ ordinary TQ, including delayed library conversion; supported grouped drafting also supports delayed library conversion |
+| `--kv-quant off` / unset | ✅ bf16 (the L1 default) | ✅ bf16 |
+| `--paged-kv` | ✅ | ✅ Gemma4 bf16; prompt-cache bypass |
+| `--memory-budget` | ✅ per-request admission | ✅ per-request admission — not aggregate (use `--kv-budget`) |
+| `--kv-budget` | n/a | ✅ aggregate queue/reject across rows |
+| `--prompt-cache` / `--ssd-cache` | ✅ prefix reuse + SSD restore | ✅ on both lanes: joiners `take()` at admission; never-merged rows `put()` back |
 | `--temperature`/`--top-p`/`--top-k` | ✅ | ✅ (per-row) |
 | `--thinking` | ✅ | ✅ |
-| vision request | ✅ | ✅ via serial lane (in bf16 under Option B) |
-| LoRA `adapter` | ✅ | ✅ via serial lane |
-| `repetition_penalty` | ✅ | ✅ (batches — per-row logits processors, since 2026-07-02) |
-| `min_p` / `xtc_*` / `logit_bias` / presence+frequency penalties | ✅ | ✅ (batches — same per-row processors) |
-| `seed` | ✅ | ✅ via serial lane |
+| vision / audio / video request | ✅ | ✅ via serial lane |
+| LoRA `adapter` | ✅ | ✅ compatible groups on adapter-capable backends |
+| `repetition_penalty` / `min_p` / `xtc_*` / `logit_bias` / presence+frequency | ✅ | ✅ (batches — per-row processors) |
+| `seed` | ✅ | ✅ request-local sampling |
 | `tools` / `stop` | ✅ | ✅ (batches) |
 | structured output (`response_format`/`guided_*`) | ✅ (mask in the decode loop) | ✅ (batches; per-row matchers) |
-| `--draft-model` | ✅ spec decode (grammar composes) | ⚠️ mounts, but routes **every** request serial — spec and batching are different modes |
+| `--draft-model` / `--draft-kind` | ✅ eligible spec decode | ✅ qualified MTP/lookup/standalone-draft groups, including grammar/logprobs; other providers remain serial |
 | GLM `--mtp on` | ✅ native MTP spec decode | ⚠️ default-on MTP routes every request serial+spec; `--mtp off` exposes ordinary GLM batching |
-| `--compiled-decode` | ✅ (serial decode route) | ✅ at **B=1 only** (Phase 3.2): a lone request's adopted serial-class caches replay the same compiled step, same kill switch; B>1 steps run the plain graph |
-| `--fused-sdpa`/`--force-wire` | ✅ (serial decode route) | **n/a — compat mode, no perf flags by design** |
+| `--compiled-decode` | ✅ | ✅ at **B=1 only**: a lone request's adopted serial-class caches replay the same compiled step; B>1 steps run the plain graph |
+| `--fused-sdpa` / `--force-wire` | ✅ (serial decode route) | n/a — compat mode, no perf flags by design |
 
 ### `--batch N` is compat mode — perf flags don't apply by design
 
 The bit-parity guarantee (mlx-lm B=N) is the *whole point* of `--batch N`,
-and it requires running the plain forward path. So the batched lane
-deliberately exposes **no** perf knobs: the scheduler
+and it requires running the plain forward path. The scheduler
 ([batch-scheduler.ts](../../src/serve/batch-scheduler.ts)) drives the
 model through `forwardHidden`/`logitsFromHidden` directly (not
-`generate()`), running the same bit-exact kernels mlx-lm runs — never the
-optional, parity-breaking ones. Flagging the batched lane would defeat
-the guarantee, so it's intentionally not wired.
-
-- **`--compiled-decode`** — engages at **B=1 only** (Phase 3.2,
-  adopt-don't-copy): a lone request's caches stay serial-class, so the
-  scheduler replays the serial engine's compiled step — bit-exact by the
-  same gate, `MLX_BUN_COMPILED_DECODE=0` is the same kill switch. The
-  moment a second row joins, steps run the plain batched graph.
-- **`--fused-sdpa`** — never engages in the batched lane (the fused
-  quantized-KV SDPA is an L2 serial-lane composition). See the
-  validation matrix in [parallel-slots.md](../design/parallel-slots.md).
-- **`--force-wire`** — doesn't wire (the scheduler bypasses `generate()`'s
-  wired scope). A model that needs wiring for speed (e.g. 26B: 8.6 → 32.3
-  tok/s wired) runs at mlx-lm-equivalent unwired speed under `--batch N`;
-  in practice such a model has no headroom for B>1 KV anyway.
-- **Always-on bit-exact kernels still run.** The compiled activations
-  (mlx-lm's own `@mx.compile` geglu/swiglu) are bit-exact with the
-  spelled-out MLP, so they stay on in both lanes without breaking
-  parity — "compat mode" means *no parity-breaking optionality*, not
-  -O0.
-
-The serial lane is where the decode route lives: `--compiled-decode` /
-`--fused-sdpa` / `--kv-quant` (and the `--l2` optiq composition) engage
-there.
+`generate()`), so:
+
+- **`--compiled-decode`** engages at **B=1 only** (adopt-don't-copy: the
+  lone request's caches stay serial-class, so the scheduler replays the
+  serial engine's compiled step — same kill switch). A second row ⇒ the
+  plain batched graph.
+- **`--fused-sdpa`** never engages in the batched lane (an L2 serial-lane
+  composition).
+- **`--force-wire`** doesn't wire (the scheduler bypasses `generate()`'s
+  wired scope).
+- **Always-on bit-exact kernels still run.** The compiled activations are
+  bit-exact with the spelled-out MLP, so they stay on in both lanes —
+  "compat mode" means *no parity-breaking optionality*, not -O0.
 
 ## Known limitations under `--batch N`
 
-These are deliberate v1 scope, not bugs — but they change behavior, so
-know them:
-
-1. **Prompt cache on the SPEC path is bypassed.** A `--draft-model`
-   server re-prefills every request; the target+draft cache-entry
-   composition is designed in mlx-lm-tool-parity-plan §7.6, not built.
-   (Batched rows DO reuse the prompt cache since Phase 3.2 — see the
-   lane table above and `/stats` `submitted_rows`.)
-2. **Aggregate admission is opt-in.** `--memory-budget` checks each
-   request against single-sequence max-safe-context; the AGGREGATE cap
-   across N concurrent rows is `--kv-budget <GB>` (landed 2026-07-03):
-   over-budget joiners queue until rows finish, a request over the budget
-   alone rejects. Without `--kv-budget`, N large-context rows can still
-   collectively exceed memory (uncatchable GPU OOM) — set it when running
-   `--batch N` near the machine's limit.
-3. **Short-context only.** Verified pre-ring-wrap (rows < the 1024
-   sliding window). Long-context (context > window) batched decode is a
-   separate validation.
-4. **bf16 by default — by contract; mixed-KV batching beyond it.**
-   mlx-lm's batched path *is* bf16, so bf16 batching is exactly what
-   mlx-lm-parity means. Phase 3.1 adds batched per-layer MIXED KV for
-   all-full-attention configs — a beyond-mlx-lm composition, verified
-   per row against the optiq oracle instead (not a
-   shortcoming). Going further — batched + mixed-precision KV quant — is
-   novel territory with no mlx-lm/optiq oracle, so it's a deferred,
-   KL-gated extension.
-5. **`extend` join (landed 2026-07-03).** A joining request appends to
-   the running batch's full-attention KV in one pad+concat (mlx-lm
-   `BatchKVCache.extend` semantics, token-exact vs its oracle);
-   `MLX_BUN_BATCH_EXTEND=0` reverts to the old whole-batch re-merge
-   (numerically equivalent, O(B·S)). Sliding-window layers still re-merge
-   on join (rotating-extend is a follow-up).
+Deliberate v1 scope, not bugs:
+
+1. **Other draft providers lack shared state.** Qwen MTP, prompt lookup and
+   standalone drafting reuse aligned prefill and completed-output state through
+   the common RAM/SSD cache. Supported full-attention targets also retain
+   lookup and standalone-draft state. Gemma assistant drafting retains its target hidden
+   and donor KV through the same cache. Other providers start fresh. Long-conversation timing and pressure
+   acceptance remain open.
+2. **Aggregate admission is opt-in** via `--kv-budget`; without it N
+   large-context rows can collectively exceed memory.
+3. **Coverage depends on composition.** Same-B Gemma oracle tests cover
+   sliding-window wrap and late joins. Broader media, resume, paged and method
+   combinations remain open in Phase 18.
+4. **bf16 by contract; mixed-KV batching beyond it.** mlx-lm's batched
+   path *is* bf16. Per-layer `config` batching is a beyond-mlx-lm
+   composition verified per row against the optiq oracle. Batched
+   uniform/TurboQuant now has shared layouts; additional same-B external
+   oracle and feature-composition checks remain.
+5. **`extend` join** appends a joining request to the running batch's
+   full-attention KV in one pad+concat (`MLX_BUN_BATCH_EXTEND=0` reverts
+   to whole-batch re-merge); sliding-window layers still re-merge on join.
 
 ## Fidelity tiers and the decode route (`--l1` / `--l2`)
 
-The tiers are **correctness contracts**, and each flag is an alias for a
-decode-route preset (`applyDecodeRoute()` in [src/cli.ts](../../src/cli.ts));
-any per-fork flag (`--kv-quant`, `--compiled-decode`,
-`--compiled-activations`, `--fused-sdpa`)
-overrides its tier's preset.
+The tiers are **correctness contracts**; each flag is an alias for a
+decode-route preset (`applyDecodeRoute()` in `src/cli.ts`), and any
+per-fork flag (`--kv-quant`, `--compiled-decode`, `--compiled-activations`,
+`--fused-sdpa`) overrides its tier's preset.
 
 | Tier | Contract | KV | Kernels | Verified against |
 | --- | --- | --- | --- | --- |
-| `--l1` | mlx-lm **bit-for-bit** | bf16 | compiled-decode + compiled activations (faithful geglu/swiglu) | mlx-lm goldens (per machine) |
-| `--l2` | mlx-optiq **bit-for-bit** | mixed-precision (`kv_config.json`) | + fused-SDPA (matches optiq exactly) | optiq goldens |
-| *(none)* | **= `--l1`** | bf16 | compiled-decode + compiled activations | mlx-lm goldens |
-
-Compiled-decode is on in **every** tier (proven bit-exact with uncompiled —
-free speed, not a fidelity trade).
-
-**There is no `--l3` tier anymore** (removed 2026-07-05; passing `--l3`
-errors out with a pointer). The tier had no content: every kernel it
-gated lost its paired A/B against the L1 baseline and was deleted. Its
-role is replaced by the **Lab** — output-changing experiments run as env
-flags with a bench + expiry, and one graduates only by beating the L1
-baseline in a paired A/B. See
-[docs/design/unified-engine-frontier-plan.md](../design/unified-engine-frontier-plan.md).
-
-**KV precision is a separate axis from the tier's kernels.** Each tier row shows its
-*default* KV scheme, but KV is not what defines the tier. mlx-lm supports **uniform**
-quantized KV (`--kv-bits`), and `--l1 --kv-quant 8` is a **bit-exact L1** config: with
-fused-sdpa off (the L1 default) our quantized decode runs `quantizedSdpaUnfused`,
-which is op-for-op identical to mlx-lm's `quantized_scaled_dot_product_attention`
-(`mlx_lm/models/base.py`) — same `mx.quantized_matmul` ×2 + `mx.softmax(precise=True)`
-+ `where(…, finfo.min)`. bf16 is just the simplest L1 default. The **fused / N-tiled**
-quantized path (fused-sdpa ON — the `--kv-quant config` / `--l2` composition) is the
-**optiq**-aligned one. Only the **per-layer mixed-precision** `config` scheme is
-optiq-only (→ L2).
+| `--l1` | mlx-lm **bit-for-bit** | bf16 | compiled decode + compiled activations, fused-sdpa off | mlx-lm goldens (per machine) |
+| `--l2` | mlx-optiq **bit-for-bit** | mixed precision (`kv_config.json`) | + fused SDPA (matches optiq exactly) | optiq goldens |
+| *(none)* | **= `--l1`** | bf16 | as `--l1` | mlx-lm goldens |
+| **Lab** | no external oracle | — | env-flagged experiments with a bench + expiry; graduate only by beating L1 in a paired A/B | validity / KL / eval gates |
+
+Compiled decode is on in **every** tier (proven bit-exact with uncompiled —
+free speed, not a fidelity trade). **There is no `--l3` tier** (removed
+2026-07-05; its role is the Lab).
+
+**KV precision is a separate axis from the tier's kernels.** `--l1
+--kv-quant 8` is a bit-exact L1 config: with fused-sdpa off our quantized
+decode runs `quantizedSdpaUnfused`, op-for-op mlx-lm's
+`quantized_scaled_dot_product_attention` (`mlx_lm/models/base.py`). Only
+the per-layer `config` scheme is optiq-only (→ L2).
 
 **Where each feature sits:** batching = L1-class (mlx-lm B=N parity);
-speculative decoding = L1-class (token-for-token vs mlx-lm's spec path);
-structured output = L2-class (oMLX oracle; masking doesn't touch the
-numerics of valid tokens, so it composes with any tier); grammar × spec
-together, HLG sampling, expert offload, and (future) batched quantized KV
-are **no-oracle (Lab-gated)** — no ancestor runtime does them, so they're
-gated by validity/KL/quality instead of bit-parity.
+two-model speculative decoding = L1-class; structured output = L2-class
+(oMLX oracle; masking doesn't touch the numerics of valid tokens, so it
+composes with any tier); grammar × spec, HLG sampling, expert offload,
+TurboQuant's cache class, jump-forward decoding, and batched
+uniform/turbo KV are **no-oracle (Lab-gated)**.
+
+## Feature matrix
+
+Everything mlx-bun serves, with its default, lane, fidelity tier, and knob.
+
+### Serving & performance
+
+| Feature | Default | Lane | Tier | Knob |
+| --- | --- | --- | --- | --- |
+| OpenAI chat completions (`/v1/chat/completions`, SSE) | on | both | — | — |
+| Raw text completions (`/v1/completions`) | on | both | — | — |
+| Anthropic Messages (`/v1/messages`) | on | both | — | — |
+| OpenAI Responses shim (`/v1/responses`) | on | both | — | — |
+| Continuous batching (mlx-lm parity at the same B) | on, cap 8 | batch | L1 | `--batch <n>` (`1` pins serial) |
+| Prompt cache (prefix KV reuse) | on, 8 GiB | both | — | `--prompt-cache <GB>` (`0` = off) |
+| SSD KV cold tier (survives eviction + restarts; GLM compressed MLA/DSA state included) | off | both | — | `--ssd-cache <dir>` (+ `--ssd-cache-max`, `--ssd-cache-verify`, `--ssd-demote-idle`, `--generation-checkpoint`) |
+| Runtime isolation (crash-isolated engine child behind a proxy parent) | off | both | — | `--isolate` |
+| Model pool (LRU-capped resident engines under `--isolate`) | 1 | both | — | `--model-pool <n>` |
+| Mixed-precision KV (`kv_config.json`, optiq's scheme) | off | serial + batch (per-layer configs batch) | L2 | `--kv-quant config`, `--l2` |
+| Uniform quantized KV (mlx-lm's `--kv-bits` scheme) | off | serial + batch | L1 | `--kv-quant 4\|8` |
+| TurboQuant KV (rotation-based: affine keys + FWHT/Lloyd-Max values) | off | serial + qualified shared ordinary/drafting | Lab (codec oracle-backed; composition gates apply) | `--kv-quant turbo[:k<bits>v<bits>]` (default `k8v3`) |
+| Compiled decode (bit-exact graph replay) | on | serial, batch at B=1 | L1/L2 | `--compiled-decode on\|off` |
+| Compiled activations (mlx-lm's `@mx.compile` geglu/swiglu) | on | both | L1 | `--compiled-activations on\|off` |
+| Fused SDPA (optiq-exact quantized-KV attention) | follows `--kv-quant` | serial | L2 | `--fused-sdpa on\|off` |
+| Paged KV (vLLM-style block pool, gather before the stock SDPA) | off | shared B1/B>1 | gated bit-exact vs plain `KVCache` | `--paged-kv`, `--paged-kv-block-size <n>` |
+| Speculative decoding (two-model / Gemma `-assistant` / DSpark / DeepSpec / native MTP head) | off | shared Qwen methods; full-attention/rotating lookup/standalone and Gemma assistant/DeepSpec/DSpark; other combinations serial | per-drafter oracle (L1 two-model, L2 assistant, DeepSpec reference, Lab DSpark) | `--draft-model`, `--draft-kind`, `--num-draft-tokens` |
+| GLM-5.2 checkpoint-native MTP | on for GLM-5.2 | serial+spec | oracle trajectory + synthetic HTTP gate | `--mtp on\|off` |
+| Model-free prompt-lookup speculation (vLLM `ngram` port) | off | Qwen and full-attention targets shared / other models serial | lossless by verify (gated vs non-spec greedy) | `--draft-kind ngram`, `--ngram-max`, `--ngram-min`, `--num-draft-tokens` (default 10) |
+| Estimated memory admission | off (advisory) | both | — | `--memory-budget <GB>` opts in |
+| Aggregate KV admission for batch rows | off | batch | — | `--kv-budget <GB>` |
+| Expert offload (MoE experts on mmap) | off | serial | Lab | `--expert-offload` |
+| Extend-join (O(1) batch admission) | on | batch | L1 (mlx-lm `extend`) | `MLX_BUN_BATCH_EXTEND=0` |
+| Vectorized greedy batch sampling | on | batch | bit-equal A/B | `MLX_BUN_BATCH_VEC_SAMPLE=0` |
+| Pipelined batch token readback | on | batch | numerically equivalent | `MLX_BUN_BATCH_NO_PIPELINE=1` |
+
+### Request features
+
+| Feature | Default | Lane | Tier | Knob |
+| --- | --- | --- | --- | --- |
+| Structured output (`response_format` json_object/json_schema) | on | both | L2 (oMLX) | request field; `MLX_BUN_GRAMMAR=0` kills |
+| Structured-output jump-forward (grammar-forced spans in one forward) | off | serial only | Lab | `MLX_BUN_GRAMMAR_JUMP=1` |
+| Token fast-forwarding for tool calls (template-determined spans in one forward) | off | serial only | opt-in; identity tested on covered fixtures, parser/held-out gates remain (`tests/parity/fill-strict.test.ts`) | `MLX_BUN_FILL=strict` |
+| Echo injection (session self-copy spans, verified against the same forward's logits) | off | serial only | Lab (paired A/B on task success + wall clock before any default) | `MLX_BUN_FILL=echo`, `MLX_BUN_FILL_K`, `MLX_BUN_FILL_CANDIDATES`, `MLX_BUN_FILL_INDEX_MAX` |
+| `guided_grammar` (EBNF) / `guided_regex`¹ / `guided_choice` / `structured_outputs` | on | both | L2 | request fields |
+| Structured output × speculative decoding | on when both active | Qwen grouped MTP, lookup and standalone drafting / eligible serial methods | Lab | — |
+| Quantized KV × speculative decoding | Shared Qwen supports uniform KV4/KV8 and TQ with zero or positive conversion thresholds; strict serial supports qualified KV4; per-layer mixed KV remains excluded | Qwen grouped MTP, lookup and standalone drafting / eligible serial methods | scheme and method gates | `MLX_BUN_QWEN_SPEC_KV4=0` disables the KV4 composition |
+| Tool calling (Gemma sentinel / CPM+Qwen XML / GLM `arg_key`+`arg_value`) + `role:"tool"` loops | on | both | — | request `tools` |
+| Vision (`image_url`; PNG/JPEG/HEIC/AVIF/WebP/TIFF/GIF/BMP) | on for models with a tower; SSRF guard on remote URLs | serial | L1/L2 | `--allow-private-media` |
+| Video input (`video_url`/`video`; AVFoundation sidecar, 2 fps, ≤768 frames, 256 MB body cap; never with audio) | on for Qwen3.5-family | serial | mlx-vlm oracle | `--allow-private-media`, `MLX_BUN_FRAME_EXTRACT` |
+| Audio input (`input_audio`/`audio`/`audio_url`; WAV native, mp3/m4a/flac/ogg/aiff via CoreAudio; ≤30 s per clip; mixes with images) | on for models with `audio_config` + sidecar tower (e4b) | serial³ | L2 (greedy stream exact vs optiq's internal model) | `--allow-private-media` |
+| LoRA adapters (mount at start / hot-swap) | off | compatible groups | — | `--adapter <dir>`, `POST /v1/adapters` |
+| Sampling: temperature / top-p / top-k / min-p / XTC / logit_bias / presence+frequency+repetition penalties | per request | both | L1 (mlx-lm-faithful) | request fields / server defaults |
+| `logprobs` / `top_logprobs` | off | both | L1 | request fields |
+| Fixed `seed` reproducibility | off | both | — | request field; compare the same execution composition |
+| Thinking-mode control (hybrid-reasoning models) | model default | both | — | `--thinking`, `chat_template_kwargs`, `reasoning_effort` |
+| Stop sequences / streaming / usage accounting | on | both | — | request fields |
+| HLG tone-curve sampling | off | serial | Lab | `--hlg-sampling on` |
+| Spec-decode telemetry (`usage.speculation`) | on with a draft | Qwen shared MTP, lookup and standalone drafting / eligible serial methods | — | — |
+| Token-fast-forwarding telemetry (`usage.fill`) | on with `MLX_BUN_FILL=strict` | serial | — | — |
+| Per-turn lane telemetry (`usage.lane`: serial / serial+spec / batched) | on | both | — | — |
+
+### Model coverage (per-model validated cells)
+
+| Family | Serial | Batch | Notes |
+| --- | --- | --- | --- |
+| MiniCPM5 (cpm5) | ✅ L1/L2 | ✅ | the starter model |
+| Gemma 4 (1B/e4b/12B/26B, + vision e4b/12B, + audio e4b) | ✅ L1/L2 | ✅ | sliding+full interleaved; MoE 26B |
+| Qwen3.5 (gated-DeltaNet hybrid) | ✅ L1/L2 | ✅ (SSM path) | `MLX_BUN_BATCH_SSM=0` reverts |
+| Qwen3.8-27B (same qwen3_5 graph) | ✅ L1 | ✅ (SSM path) | native MTP head via `--draft-model`/`--draft-kind mtp` (lossless-gated; slower on a quiet box — opt-in); images and video serve (mlx-vlm oracle) |
+| GLM-5.2 / Colibri | ✅ chat/text, Messages, Responses, SSE, tools, grammar, logprobs | ✅ compressed MLA/DSA scheduler | native MTP defaults to serial+spec; `--mtp off` enables batching; embeddings, vision/audio, adapters, training unsupported |
+| DiffusionGemma-26B (non-autoregressive) | ✅ (own engine) | — serial always | first bit-exact non-AR port |
+| Tier-0 universal (llama/qwen2/qwen3/olmo2/…, 11 archs) | ✅ L1 | ✅ plain full-attention archs² | gemma2-family / sliding-window universal → serial |
+
+¹ `guided_regex` accepts the regex∩EBNF subset today (no `\d`/anchors — those degrade to prompt injection).
+² Gated token-exact vs mlx-lm B=2 (static + dynamic join/leave) on Llama-3.2-3B.
+³ Audio is a capability neither ancestor serves (mlx-lm strips the tower; optiq never wires it into its serve frontend). Serial by design; prompt cache skipped.
+
+Beyond serving, the same binary does training (`mlx-bun train`),
+quantization (`convert`, `fuse`), embeddings, local memory (`mlx-bun
+memory`), the pi agent (`mlx-bun pi`), the registry/`fit`/`gc`, HF
+`upload`, and the web Model Hub — see [cli.md](./cli.md).
 
 ## Performance characteristics & recipes
 
 Reference numbers are from this project's dev machines (loaded-machine
-numbers are directional only; `benchmarks/RESULTS.md` holds the quotable
-set, and `scripts/bench-feature-matrix.ts` measures the six composition
-cells in one run).
-
-- **Serial is the fastest single stream** — prompt cache, mixed-precision
-  KV, compiled decode. The batch lane at ONE live row is ~25% slower
-  (cpm5 ~149 vs 193 tok/s: no prompt cache, no kv-quant, wrapper
-  overhead), which is why serial stays the default.
+numbers are directional only; [benchmarks.md](./benchmarks.md) holds the
+quotable set, and `scripts/bench-matrix.ts features` measures the
+composition cells in one run).
+
+- **A lone request is the serial engine** — prompt cache, compiled
+  decode, and SSD restore all apply at B=1 under the default cap.
 - **`--batch N` wins under concurrency** — cpm5 `--batch 4`: ~349 tok/s
-  aggregate vs ~173 serial-queued, TTFT 2–3× better; Llama-3B at just
-  B=2: 1.7× aggregate, TTFT 765→162 ms.
+  aggregate vs ~173 serial-queued, TTFT 2–3× better; Llama-3B at B=2:
+  1.7× aggregate, TTFT 765→162 ms.
 - **Spec pays only on slow targets.** One accepted draft = one skipped
-  target forward, so value scales with target step cost: 12B ≈ 1.09× at
-  γ=1 (measured); fast small targets LOSE (e4b 0.78×; 3B smoke: spec
-  61.5 vs serial 81.5 tok/s). Draft for 12B+, skip below.
-- **Structured output is ~free** (<1% serial; the batch lane drops to
-  read-before-build scheduling while a grammar row is live — bounded,
-  ~0.1 ms/step class).
-- **`--ssd-cache` has 0% decode overhead** — pure TTFT/restart win; no
-  reason to leave it off on a persistent server.
+  target forward: 12B ≈ 1.09× at γ=1; fast small targets lose (e4b
+  0.78×). Draft for 12B+, skip below.
+- **Structured output is ~free** (<1% serial; bounded ~0.1 ms/step class
+  in the batch lane).
+- **`--ssd-cache` has 0% decode overhead** — pure TTFT/restart win.
 - **Quantized-KV prefill pays a scheme-intrinsic tax at long context** —
-  chunked prefill converts each chunk's KV at the chunk boundary, so
-  every later chunk attends against the already-quantized prefix
-  (quantize-on-write; the same streaming conversion optiq serve uses —
-  `maybeQuantizeKv` in `src/generate.ts` is the port). That cost is by
-  design, not an implementation gap: ~30% prefill throughput vs bf16 at
-  ~16k where the config quantizes every cache (the 1B starter measured
-  −33% in the 2026-07-06 serve pass), single-digit % on the
-  sliding-window gemmas (rotating caches stay window-bounded). The lever
-  is upstream — mlx's `quantized_matmul` kernels (the split-K work,
-  [ml-explore/mlx#3120](https://github.com/ml-explore/mlx/pull/3120);
-  status in [decode-speed-program.md](../design/decode-speed-program.md)
-  lever 2) — not this codebase.
+  chunked prefill converts each chunk's KV at the boundary, so later
+  chunks attend against the quantized prefix (the same streaming
+  conversion optiq serve uses): ~30% prefill throughput vs bf16 at ~16k
+  where the config quantizes every cache, single-digit % on the
+  sliding-window gemmas. The lever is upstream (mlx `quantized_matmul`,
+  [decode-speed-program.md](../design/decode-speed-program.md) lever 2).
 
 **Recipes:**
 
 - *Single-user agent/chat (the default use):*
-  `mlx-bun serve <model> --ssd-cache <dir>` — L1 serial, prompt cache +
-  SSD tier. On 12B+ add `--draft-model <small-same-tokenizer>`.
+  `mlx-bun serve <model> --ssd-cache <dir>` — L1, prompt cache + SSD tier.
+  On 12B+ add `--draft-model <small-same-tokenizer>`.
 - *Several clients at once (throughput):*
-  `mlx-bun serve <model> --batch 4 --ssd-cache <dir>` — don't set
-  `--kv-quant` (it would just un-batch everything), don't mount a draft.
-- *Reproducibility:* bare / `--l1` (≡ mlx-lm), `--l1 --batch N`
-  (≡ mlx-lm B=N), `--l2` (≡ optiq).
-- *Memory-tight big model:* serial + `--kv-quant config|4|8` +
+  `mlx-bun serve <model> --batch 4 --ssd-cache <dir> --kv-budget <GB>` —
+  supports qualified start-zero TurboQuant and a grouped Qwen MTP companion.
+  Other draft providers still use serial speculation.
+- *UI must never lag / survive engine crashes:* add `--isolate`
+  (`--model-pool 2` to keep two models resident).
+- *Reproducibility:* bare / `--l1` (≡ mlx-lm), `--l1 --batch 1` (strict
+  serial), `--l2` (≡ optiq).
+- *Memory-tight big model:* `--kv-quant config|4|8|turbo` +
   `--memory-budget <GB>` + `--ssd-cache <dir>`; MoE adds
-  `--expert-offload`. Inherently the serial recipe until batched
-  quantized KV lands.
-
-**The two exclusions to remember:** batching excludes uniform/TurboQuant KV
-(per-layer `kv_config.json` does compose with batching), and spec excludes
-prompt-cache reuse (v1 bypass; the §7.6 composition is the fix).
+  `--expert-offload`. Start-zero TurboQuant supports shared execution.
+
+**Remaining exclusions:** media, paged KV, generation resume, fill and draft
+providers without grouped implementations still need shared execution work.
+Qwen ordinary, shared MTP and prompt lookup support positive library thresholds for uniform
+affine KV4/KV8 and TurboQuant. Ordinary Gemma supports positive affine KV4/KV8
+thresholds, including sliding-window layers. Other-model delayed affine remains
+open. Qwen MTP
+prefill and completed decode state use the common RAM/SSD cache. Generated
+checkpoints contain only processed tokens; persistence uses the existing
+queue. Strict serial deletion and complete feature parity
+remain acceptance work in Phase 6/18.
 
 ## Observability — `GET /stats`
 
 The live config and batch state. The canonical field-by-field snippet
-lives in [server-api.md](server-api.md#get-stats) (kept in lockstep with
-the code — this section only explains how to read it); highlights:
-`prompt_cache` (8 GB default cap), `kv_quant.mode` (incl. `turbo kXvY`),
-the conditional `ssd_cache` block, and `batch`
+lives in [server-api.md](server-api.md#get-stats); highlights:
+`prompt_cache` (cap), `kv_quant.mode` (incl. `turbo kXvY`), the
+conditional `ssd_cache` block (pending/dropped/failed spill counters), the
+conditional `glm52` block, and `batch`
 (`mode`/`active_rows`/`pending_rows`/`submitted_rows`/`kv_bytes`/
 `kv_budget_bytes`).
 
 `batch.mode` is the truthful configured/capability state: `off` for
 `--batch 1`, `serial` when a larger configured cap cannot batch the loaded
-model's cache layout, and `batch` when the model is admitted. `batch.batched`
-is the compatibility boolean for `mode == "batch"`; `active_rows` reports the
-instantaneous live rows.
+model's cache layout, and `batch` when the model is admitted.
+`batch.batched` is the compatibility boolean for `mode == "batch"`;
+`active_rows` is the instantaneous live-row count. Under `--isolate`,
+`GET /engine` on the parent reports the child pid, restart count, socket,
+and pool residency.
+
+Managed quantization and finetuning jobs wait for the serving gateway to drain,
+then hold its execution lease until their subprocess exits. Generation and native
+preparation wait during that job. Shutdown cancels queued jobs and terminates the
+host's active job before releasing its lease. Resident model weights remain
+loaded. Under isolation, the parent coordinates all of its model-pool workers.
+This controls execution, not memory capacity or independently launched servers.
+
+### Delayed KV conversion in library requests
+
+The CLI keeps `quantizedKvStart=0`. A library `KvScheme("turbo", ...)` can
+set a positive threshold for ordinary or Qwen MTP shared execution. Every row converts
+at its own existing maintenance boundary. During the mixed phase, the
+cache retains plain and encoded rows independently while model attention
+still runs at the shared batch size. The fetched attention state restores
+only the rotated value rows; once all rows convert, the standard packed
+batch layout takes over. Shared MTP converts committed history before a
+verification round; crossing the threshold inside that round converts at
+resolution after rollback and before checkpoint publication. Rejected tokens
+cannot set the conversion offset.
+
+Qwen ordinary, shared MTP and prompt lookup also support positive `quantizedKvStart` with
+uniform affine KV4/KV8. Affine and TQ adapters share the same row-transition
+lifecycle. During mixed affine precision, each row uses its existing plain or
+quantized attention arithmetic; after conversion, the packed batch layout owns
+attention. The model binding declares this support; scheduling does not choose
+the codec or conversion boundary. Ordinary Gemma also supports delayed affine
+conversion in full and sliding-window layers. Rotating rows retain the same
+physical columns while converting independently, so shared attention masks stay
+aligned. Gemma speculative methods and other-model delayed affine remain open.
+
+A delayed conversion records the earliest reusable prefix offset. RAM and
+SSD reuse preserve earlier plain donors and never trim converted state
+below that boundary. Older TQ SSD files without this metadata remain
+usable at their stored offset, but cannot serve shorter prefixes. The
+start-zero default-group cache keys are unchanged; positive TQ and affine
+thresholds, and non-default affine group sizes, have distinct keys.
+No memory budget or admission default changes.
+
+
+Supported grouped draft providers also accept TurboQuant with positive library
+conversion thresholds. Full-attention layers use the selected codec; sliding
+layers stay bf16. Generated target and method-companion snapshots preserve
+precision through RAM and queued SSD persistence. Defaults are unchanged.
+Composition evidence and remaining gaps: [batching design](../design/batching.md).
+
+The programmatic `createServer` option `quantizedKvStart` sets the absolute
+per-row offset at which the selected KV scheme converts. Omission preserves
+the existing start-zero server policy. This option is resolved once with the
+KV scheme and shared by request preparation, execution, and checkpoint identity.
+Generation checkpoints preserve encoded state and each cache's earliest reusable
+offset; restoring a checkpoint does not re-quantize it. TurboQuant retains its
+existing full-attention scope, with sliding-window caches in bf16.