opencode-cmd-provider 1.6.1 → 1.6.3

package/CHANGELOG.md

@@ -1,4 +1,88 @@
-# Changelog
+## 1.6.3 - 2026-09-02
+
+### Model catalog
+
+## Model catalog
+
+- **FACTS_LAST_REFRESHED**: `2026-09-01` → `2026-09-01`
+
+No changes.
+
+### Reasoning classification
+
+## Reasoning classification
+
+- **CLASSIFICATION_LAST_REFRESHED**: `` → `2026-09-02`
+
+_first refresh — no classification data on the before side._
+
+### Deals intelligence
+
+## Deals intelligence
+
+- **DEAL_LAST_REFRESHED**: `2026-09-01` → `2026-09-01`
+
+No changes.
+
+### Full changeset
+
+The generated-catalog diff above is only the data layer (the new classification module — first refresh). This PR carries the whole derived-classification implementation (#109–#116): the shared RSC record source (scripts/rsc-source.mjs), the classification generator in the refresh ladder, the derived runtime (src/provider/reasoning.ts), the semantic diff sections + classificationChanged boolean, the meaningful-change drift detection (date-only churn opens no PR), ADR-0006/0007 + glossary + docs, and the auto-release workflow. npm test green on the head commit; per-ticket summary in the commit trail.
+
+### Maintainer actions before merge (ADR-0007)
+
+1. Add the RELEASE_PUSH_TOKEN secret (fine-grained PAT, this repo only, Contents: Read and write) so the tag push triggers the release pipeline — a default-token push does not start other workflows.
+2. Branch protection is currently off, so the bot push lands freely; when you enable it, add a bypass for the pushing identity (ruleset bypass list).
+
+_Opened from `catalog-refresh/2026-09-02` so merging exercises the new merge-to-release path: the auto-release workflow cuts `chore(release): 1.6.3` + tags v1.6.3, and release.yml publishes._
+
+## 1.6.2 - 2026-09-01
+
+Patch release. Make the daily catalog-refresh cron self-sustainable against
+upstream data changes (refs issue #89) and ship the 2026-09-01 catalog
+refresh.
+
+- **Self-sustainable cron tests** — the daily catalog-refresh cron has been
+  failing every day since 2026-08-28 because its shape-pinning tests pinned
+  upstream-managed data (deal pcts, benchmark throughput, plan allowances,
+  effort-classification lists). Any legitimate upstream change — a deal
+  expiring, a benchmark value moving, a free variant being retired — failed
+  `npm test` and blocked the refresh PR until a human updated the pins.
+  This release separates **shape** from **value** in those tests, so the
+  cron can land upstream changes without manual intervention:
+  - `tests/refresh-deals.test.ts`: the five real-RSC `modelDealEntry`
+    value pins (Gemini 3.7 Flash, Qwen 3.6 Plus, MiniMax M3, DeepSeek V4
+    Flash, Laguna) are replaced with nine synthetic-record tests that own
+    every value they assert. A companion test exercises the "deal ended"
+    case the cron hit on 2026-08-31. The end-to-end real-RSC smoke test
+    still runs the parser over every per-plan slug record to catch
+    shape/contract regressions, but no longer pins values.
+  - `tests/refresh-deals-rsc.test.ts`: the "emitted catalog has parity
+    with the shipped catalog" test asserts structural fields (discount
+    exists, benchmark exists, allowance has both goat/pro keys,
+    peakOffPeak exists) instead of exact MiniMax / Kimi values. Value
+    semantics now live in the synthetic `modelDealEntry` tests.
+  - `tests/catalog-metadata.test.ts`: `EFFORTS_MODELS` is now derived
+    from the auto-generated `MODEL_EFFORTS` (the npm package's
+    `models.md` is the source of truth for which models expose effort
+    levels), so a new efforts model — like `deepseek/deepseek-v4-flash-fast`
+    added 2026-08-31 — lands in `MODEL_EFFORTS` on the next refresh and
+    the reasoning-classification test follows automatically. The
+    "exactly once" gate becomes a "no model is in both `MODEL_EFFORTS`
+    and `REASONING_MODELS`" invariant (a real misclassification signal)
+    plus an advisory check that `isReasoningModel` agrees with the
+    union and respects `NON_REASONING_MODELS`.
+  - `tests/plugin-models.test.ts`: the "free variants get a `(free)`
+    suffix" test pins on `poolside/laguna-s-2.1-free` (still in the
+    snapshot) instead of the MiniMax name-collision pair retired
+    from upstream.
+  - `src/provider/reasoning.ts`: drop `minimax/minimax-m3-free` from
+    `REASONING_MODELS` — the variant was retired from upstream's
+    `models.md` alongside the free model itself, so the entry was stale.
+- **Catalog refresh — 2026-09-01** — snapshot + capability facts +
+  RSC fixtures refreshed to current upstream values (see PR #105).
+- **Issue #89** — the broader question of how to lock down shape-pinning
+  value pins remains open; this release is the cron-side half of that
+  fix. The deals-value-pin half is still tracked in #89.
 
 ## 1.6.1 - 2026-08-29
 

package/README.md

@@ -161,6 +161,11 @@ auto-registered models only.
 - Snapshot — `src/catalog/snapshot.ts` (model ids, names, context lengths) plus
   `src/catalog/facts.ts` (reasoning efforts, per-1M-token rates, input
   modalities). Regenerated from the live catalog via `npm run refresh:snapshot`.
+- Classification — `src/catalog/classification.ts` (per-model reasoning
+  capability, derived from the `reasoning` flag on the docs' RSC slug
+  records; ADR-0006). Regenerated via `npm run refresh:classification`; the
+  runtime derives its reasoning metadata from this module, so upstream
+  classification changes land as data, never as hand edits.
 - Deals — `src/deals/catalog.ts` (per-model tier, benchmarks, deal
   discounts, `was`/`now` rates, peak/off-peak windows, and GOAT/Pro monthly
   allowances for every model). Extracted from the Command Code docs' RSC
@@ -173,12 +178,16 @@ auto-registered models only.
 Run `npm run refresh` to regenerate everything at once: the snapshot and
 facts come from the live Command Code catalog, the RSC fixtures are
 re-captured from the live docs pages (`npm run refresh:fixtures`), and the
-deals catalog is regenerated from the fresh fixtures — so fixtures, catalog,
-and tests stay in lockstep. A daily GitHub Actions cron
+classification module and deals catalog are regenerated from the fresh
+fixtures — so fixtures, catalogs, and tests stay in lockstep. A daily GitHub
+Actions cron
 (`.github/workflows/catalog-refresh.yml`, 06:00 UTC, also manually
 triggerable via `workflow_dispatch`) runs the same pipeline: when upstream
 moved it opens a `chore: catalog refresh` PR whose body is the human-readable
-diff from `scripts/diff-catalog.mjs`; when nothing drifted it exits silently.
+diff from `scripts/diff-catalog.mjs` — model catalog, reasoning
+classification (flips, new reasoning models, retirements, active overrides),
+and deals sections; when nothing meaningful drifted it exits silently (a
+run whose only change is the refreshed-date stamps opens no PR).
 
 Auto-registration adds no network latency to OpenCode startup — the snapshot is
 bundled, and the plugin never contacts the Command Code API to list models. The
@@ -199,7 +208,13 @@ compatible API endpoints:
 
 ### Reasoning support
 
-Reasoning metadata is enriched only for models whose Command Code effort support is known. Supported levels are sent as the documented `reasoning_effort` request field; `off`, unsupported levels, and newly discovered models without metadata do not add reasoning fields to the request. No prompt instructions are injected.
+Reasoning metadata is derived from the generated catalogs (ADR-0006): models
+upstream flags as reasoning-capable advertise `reasoning: true`
+automatically — with explicit effort variants when the generated facts list
+effort levels, and without variants otherwise. Supported levels are sent as
+the documented `reasoning_effort` request field; `off`, unsupported levels,
+and newly discovered models without metadata do not add reasoning fields to
+the request. No prompt instructions are injected.
 
 Reasoning blocks from completed assistant turns are not replayed to Command Code in later requests; only the user-visible text and completed tool calls are sent back as history. This prevents prior private reasoning traces from interfering with reasoning on follow-up turns.
 

package/dist/src/catalog/classification.d.ts

@@ -0,0 +1,7 @@
+export declare const CLASSIFICATION_SOURCE_URLS: ReadonlyArray<string>;
+export declare const CLASSIFICATION_LAST_REFRESHED = "2026-09-02";
+export declare const CLASSIFICATION_OVERRIDES: Readonly<Record<string, {
+    capability: boolean;
+    justification: string;
+}>>;
+export declare const MODEL_REASONING_CAPABILITY: Readonly<Record<string, boolean>>;

package/dist/src/catalog/classification.js

@@ -0,0 +1,89 @@
+// src/catalog/classification.ts — GENERATED by scripts/refresh-classification.mjs. Do not edit.
+//
+// Per-model reasoning capability, derived from the `reasoning` flag on
+// the Command Code docs' RSC slug records (plans/goat + plans/pro,
+// slug-id aliasing applied, keyed by snapshot id) — the same
+// snapshot-keyed records the Deals catalog consumes. Classification is
+// upstream-managed and derived, never hand-maintained; the only human
+// seam is CLASSIFICATION_OVERRIDES, whose entries each carry a written
+// justification naming the upstream disagreement (see ADR-0006).
+// Regenerate with `npm run refresh` (offline:
+// `npm run refresh:classification -- --fixtures`).
+export const CLASSIFICATION_SOURCE_URLS = [
+    "https://commandcode.ai/docs/plans/goat",
+    "https://commandcode.ai/docs/plans/pro",
+];
+export const CLASSIFICATION_LAST_REFRESHED = "2026-09-02";
+// Active capability overrides. Empty = upstream data is truth.
+// Every entry requires a written justification naming the upstream
+// disagreement; entries are rendered into the refresh PR body.
+export const CLASSIFICATION_OVERRIDES = {};
+export const MODEL_REASONING_CAPABILITY = {
+    "claude-fable-5": true,
+    "claude-fable-5-1": true,
+    "claude-haiku-4-5-20251001": false,
+    "claude-opus-4-7": true,
+    "claude-opus-4-8": true,
+    "claude-opus-5": true,
+    "claude-sonnet-4-6": true,
+    "claude-sonnet-5": true,
+    "deepseek/deepseek-v4-flash": true,
+    "deepseek/deepseek-v4-flash-fast": true,
+    "deepseek/deepseek-v4-flash-vision-exp": true,
+    "deepseek/deepseek-v4-pro": true,
+    "google/gemini-3.1-flash-lite": true,
+    "google/gemini-3.5-flash": true,
+    "google/gemini-3.5-flash-lite": true,
+    "google/gemini-3.6-flash": true,
+    "google/gemini-3.7-flash": true,
+    "google/gemini-3.8-flash": true,
+    "gpt-5.3-codex": true,
+    "gpt-5.4": true,
+    "gpt-5.4-mini": true,
+    "gpt-5.5": true,
+    "gpt-5.6-luna": true,
+    "gpt-5.6-sol": true,
+    "gpt-5.6-terra": true,
+    "meituan/LongCat-2.0:free": true,
+    "meta/muse-spark-1.1": true,
+    "meta/muse-spark-1.2": true,
+    "meta/muse-spark-1.2-contributor": true,
+    "meta/muse-spark-1.3": true,
+    "meta/muse-spark-1.3-contributor": true,
+    "MiniMaxAI/MiniMax-M2.5": false,
+    "MiniMaxAI/MiniMax-M2.7": false,
+    "MiniMaxAI/MiniMax-M3": true,
+    "moonshotai/Kimi-K2.5": false,
+    "moonshotai/Kimi-K2.6": false,
+    "moonshotai/Kimi-K2.7-Code": true,
+    "moonshotai/Kimi-K2.7-Code-Highspeed": true,
+    "moonshotai/Kimi-K3": true,
+    "nvidia/nemotron-3-ultra-550b-a55b": true,
+    "poolside/laguna-s-2.1-free": true,
+    "Qwen/Qwen3.6-Max-Preview": true,
+    "Qwen/Qwen3.6-Plus": true,
+    "Qwen/Qwen3.7-Flash": true,
+    "Qwen/Qwen3.7-Max": true,
+    "Qwen/Qwen3.7-Plus": true,
+    "Qwen/Qwen3.8-27B": true,
+    "Qwen/Qwen3.8-Flash": true,
+    "Qwen/Qwen3.8-Max": true,
+    "Qwen/Qwen3.8-Max-0902": true,
+    "sakana/fugu-ultra": true,
+    "stepfun/Step-3.5-Flash": true,
+    "stepfun/Step-3.7-Flash": true,
+    "tencent/hy3-paid": true,
+    "tencent/hy4-preview": true,
+    "thinkingmachines/inkling": true,
+    "thinkingmachines/inkling-small": true,
+    "xai/grok-4.5": true,
+    "xai/grok-4.6": true,
+    "xiaomi/mimo-v2.5": false,
+    "xiaomi/mimo-v2.5-pro": false,
+    "z-ai/glm-5.3-flash": true,
+    "zai-org/GLM-5": false,
+    "zai-org/GLM-5.1": false,
+    "zai-org/GLM-5.2": true,
+    "zai-org/GLM-5.2-Fast": false,
+    "zai-org/GLM-5.3": true,
+};

package/dist/src/catalog/facts.d.ts

@@ -1,7 +1,7 @@
-export declare const FACTS_SOURCE_URL = "https://unpkg.com/command-code@1.38.2/dist/bundled/command-code-knowledge/reference/models.md";
-export declare const MODALITIES_SOURCE_URL = "https://unpkg.com/command-code@1.38.2/dist/cli.mjs";
-export declare const FACTS_PACKAGE_VERSION = "1.38.2";
-export declare const FACTS_LAST_REFRESHED = "2026-08-29";
+export declare const FACTS_SOURCE_URL = "https://unpkg.com/command-code@1.44.0/dist/bundled/command-code-knowledge/reference/models.md";
+export declare const MODALITIES_SOURCE_URL = "https://unpkg.com/command-code@1.44.0/dist/cli.mjs";
+export declare const FACTS_PACKAGE_VERSION = "1.44.0";
+export declare const FACTS_LAST_REFRESHED = "2026-09-02";
 export declare const MODEL_EFFORTS: Readonly<Record<string, readonly string[]>>;
 export declare const MODEL_COSTS: Readonly<Record<string, {
     input: number;

package/dist/src/catalog/facts.js

@@ -4,23 +4,27 @@
 // command-code npm package's bundled model catalog (models.md), plus input
 // modalities parsed from its CLI bundle (dist/cli.mjs). Regenerate
 // with `npm run refresh:snapshot`.
-export const FACTS_SOURCE_URL = "https://unpkg.com/command-code@1.38.2/dist/bundled/command-code-knowledge/reference/models.md";
-export const MODALITIES_SOURCE_URL = "https://unpkg.com/command-code@1.38.2/dist/cli.mjs";
-export const FACTS_PACKAGE_VERSION = "1.38.2";
-export const FACTS_LAST_REFRESHED = "2026-08-29";
+export const FACTS_SOURCE_URL = "https://unpkg.com/command-code@1.44.0/dist/bundled/command-code-knowledge/reference/models.md";
+export const MODALITIES_SOURCE_URL = "https://unpkg.com/command-code@1.44.0/dist/cli.mjs";
+export const FACTS_PACKAGE_VERSION = "1.44.0";
+export const FACTS_LAST_REFRESHED = "2026-09-02";
 export const MODEL_EFFORTS = {
     "deepseek/deepseek-v4-pro": ["high", "max"],
     "deepseek/deepseek-v4-flash": ["high", "max"],
     "deepseek/deepseek-v4-flash-vision-exp": ["high", "max"],
+    "deepseek/deepseek-v4-flash-fast": ["low", "high", "max"],
+    "moonshotai/Kimi-K3": ["low", "high", "max"],
     "z-ai/glm-5.3-flash": ["low", "high", "max"],
     "zai-org/GLM-5.3": ["low", "high", "max"],
     "zai-org/GLM-5.2": ["high", "max"],
+    "Qwen/Qwen3.8-Max-0902": ["low", "medium", "xhigh"],
     "Qwen/Qwen3.8-Max": ["low", "medium", "xhigh"],
     "Qwen/Qwen3.8-27B": ["low", "medium", "xhigh"],
     "Qwen/Qwen3.8-Flash": ["low", "medium", "xhigh"],
     "tencent/hy4-preview": ["low", "medium", "high"],
     "claude-sonnet-5": ["low", "medium", "high", "xhigh", "max"],
     "claude-sonnet-4-6": ["low", "medium", "high", "xhigh", "max"],
+    "claude-fable-5-1": ["low", "medium", "high", "xhigh", "max"],
     "claude-fable-5": ["low", "medium", "high", "xhigh", "max"],
     "claude-opus-5": ["low", "medium", "high", "xhigh", "max"],
     "claude-opus-4-8": ["low", "medium", "high", "xhigh", "max"],
@@ -32,6 +36,7 @@ export const MODEL_EFFORTS = {
     "gpt-5.4": ["low", "medium", "high", "xhigh"],
     "gpt-5.3-codex": ["low", "medium", "high", "xhigh"],
     "gpt-5.4-mini": ["low", "medium", "high"],
+    "google/gemini-3.8-flash": ["low", "medium", "high"],
     "google/gemini-3.7-flash": ["low", "medium", "high"],
     "google/gemini-3.6-flash": ["low", "medium", "high"],
     "google/gemini-3.5-flash": ["low", "medium", "high"],
@@ -45,6 +50,7 @@ export const MODEL_COSTS = {
     "deepseek/deepseek-v4-pro": { input: 0.66, output: 1.98, cacheRead: 0.022, cacheWrite: 0 },
     "deepseek/deepseek-v4-flash": { input: 0.22, output: 0.66, cacheRead: 0.007, cacheWrite: 0 },
     "deepseek/deepseek-v4-flash-vision-exp": { input: 0.22, output: 0.66, cacheRead: 0.007, cacheWrite: 0 },
+    "deepseek/deepseek-v4-flash-fast": { input: 0.28, output: 0.56, cacheRead: 0.07, cacheWrite: 0 },
     "moonshotai/Kimi-K3": { input: 3, output: 15, cacheRead: 0.3, cacheWrite: 0 },
     "moonshotai/Kimi-K2.7-Code": { input: 0.95, output: 4, cacheRead: 0.19, cacheWrite: 0 },
     "moonshotai/Kimi-K2.7-Code-Highspeed": { input: 1.9, output: 8, cacheRead: 0.38, cacheWrite: 0 },
@@ -58,11 +64,10 @@ export const MODEL_COSTS = {
     "zai-org/GLM-5": { input: 1, output: 3.2, cacheRead: 0.2, cacheWrite: 0 },
     "MiniMaxAI/MiniMax-M3": { input: 0.3, output: 1.2, cacheRead: 0.06, cacheWrite: 0 },
     "MiniMaxAI/MiniMax-M2.7": { input: 0.3, output: 1.2, cacheRead: 0.06, cacheWrite: 0 },
-    "minimax/minimax-m3-free": { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
-    "minimax/minimax-m2.7-free": { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
     "MiniMaxAI/MiniMax-M2.5": { input: 0.3, output: 1.2, cacheRead: 0.03, cacheWrite: 0 },
     "xiaomi/mimo-v2.5-pro": { input: 0.435, output: 0.87, cacheRead: 0.0036, cacheWrite: 0 },
     "xiaomi/mimo-v2.5": { input: 0.14, output: 0.28, cacheRead: 0.0028, cacheWrite: 0 },
+    "Qwen/Qwen3.8-Max-0902": { input: 2, output: 6, cacheRead: 0.25, cacheWrite: 0 },
     "Qwen/Qwen3.8-Max": { input: 2, output: 6, cacheRead: 0.25, cacheWrite: 2.5 },
     "Qwen/Qwen3.8-27B": { input: 0.4, output: 3, cacheRead: 0.04, cacheWrite: 0 },
     "Qwen/Qwen3.8-Flash": { input: 0.16, output: 0.47, cacheRead: 0.016, cacheWrite: 0 },
@@ -71,6 +76,7 @@ export const MODEL_COSTS = {
     "Qwen/Qwen3.7-Flash": { input: 0.03, output: 0.13, cacheRead: 0.006, cacheWrite: 0.038 },
     "Qwen/Qwen3.6-Max-Preview": { input: 1.3, output: 7.8, cacheRead: 0.26, cacheWrite: 1.63 },
     "Qwen/Qwen3.6-Plus": { input: 0.5, output: 3, cacheRead: 0.1, cacheWrite: 0 },
+    "meituan/LongCat-2.0:free": { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
     "stepfun/Step-3.7-Flash": { input: 0.2, output: 1.15, cacheRead: 0.04, cacheWrite: 0 },
     "stepfun/Step-3.5-Flash": { input: 0.1, output: 0.3, cacheRead: 0.02, cacheWrite: 0 },
     "tencent/hy3-paid": { input: 0.14, output: 0.58, cacheRead: 0.035, cacheWrite: 0 },
@@ -81,6 +87,7 @@ export const MODEL_COSTS = {
     "poolside/laguna-s-2.1-free": { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
     "claude-sonnet-5": { input: 2, output: 10, cacheRead: 0.2, cacheWrite: 2.5 },
     "claude-sonnet-4-6": { input: 3, output: 15, cacheRead: 0.3, cacheWrite: 3.75 },
+    "claude-fable-5-1": { input: 10, output: 50, cacheRead: 0.25, cacheWrite: 12.5 },
     "claude-fable-5": { input: 10, output: 50, cacheRead: 1, cacheWrite: 12.5 },
     "claude-opus-5": { input: 5, output: 25, cacheRead: 0.5, cacheWrite: 6.25 },
     "claude-opus-4-8": { input: 5, output: 25, cacheRead: 0.5, cacheWrite: 6.25 },
@@ -93,6 +100,7 @@ export const MODEL_COSTS = {
     "gpt-5.4": { input: 2.5, output: 15, cacheRead: 0.25, cacheWrite: 0 },
     "gpt-5.3-codex": { input: 2, output: 8, cacheRead: 0.5, cacheWrite: 0 },
     "gpt-5.4-mini": { input: 0.75, output: 4.5, cacheRead: 0.075, cacheWrite: 0 },
+    "google/gemini-3.8-flash": { input: 1.5, output: 7.5, cacheRead: 0.15, cacheWrite: 0 },
     "google/gemini-3.7-flash": { input: 1.5, output: 7.5, cacheRead: 0.15, cacheWrite: 0.08334 },
     "google/gemini-3.6-flash": { input: 1.5, output: 7.5, cacheRead: 0.15, cacheWrite: 0 },
     "google/gemini-3.5-flash": { input: 1.5, output: 9, cacheRead: 0.15, cacheWrite: 0 },
@@ -102,6 +110,8 @@ export const MODEL_COSTS = {
     "meta/muse-spark-1.1": { input: 1.25, output: 4.25, cacheRead: 0.15, cacheWrite: 0 },
     "meta/muse-spark-1.2": { input: 1.25, output: 4.25, cacheRead: 0.15, cacheWrite: 0 },
     "meta/muse-spark-1.2-contributor": { input: 0.1, output: 0.2, cacheRead: 0.002, cacheWrite: 0 },
+    "meta/muse-spark-1.3": { input: 1.25, output: 4.25, cacheRead: 0.15, cacheWrite: 0 },
+    "meta/muse-spark-1.3-contributor": { input: 0.1, output: 0.2, cacheRead: 0.002, cacheWrite: 0 },
     "xai/grok-4.5": { input: 2, output: 6, cacheRead: 0.5, cacheWrite: 0 },
     "xai/grok-4.6": { input: 2, output: 6, cacheRead: 0.5, cacheWrite: 0 },
 };
@@ -113,7 +123,9 @@ export const MODEL_INPUT_MODALITIES = {
     "Qwen/Qwen3.8-27B": ["text", "image"],
     "Qwen/Qwen3.8-Flash": ["text", "image"],
     "Qwen/Qwen3.8-Max": ["text", "image"],
+    "Qwen/Qwen3.8-Max-0902": ["text", "image"],
     "claude-fable-5": ["text", "image"],
+    "claude-fable-5-1": ["text", "image"],
     "claude-haiku-4-5-20251001": ["text", "image"],
     "claude-opus-4-7": ["text", "image"],
     "claude-opus-4-8": ["text", "image"],
@@ -126,6 +138,7 @@ export const MODEL_INPUT_MODALITIES = {
     "google/gemini-3.5-flash-lite": ["text", "image"],
     "google/gemini-3.6-flash": ["text", "image"],
     "google/gemini-3.7-flash": ["text", "image"],
+    "google/gemini-3.8-flash": ["text", "image"],
     "gpt-5.3-codex": ["text", "image"],
     "gpt-5.4": ["text", "image"],
     "gpt-5.4-mini": ["text", "image"],
@@ -136,7 +149,8 @@ export const MODEL_INPUT_MODALITIES = {
     "meta/muse-spark-1.1": ["text", "image"],
     "meta/muse-spark-1.2": ["text", "image"],
     "meta/muse-spark-1.2-contributor": ["text", "image"],
-    "minimax/minimax-m3-free": ["text", "image"],
+    "meta/muse-spark-1.3": ["text", "image"],
+    "meta/muse-spark-1.3-contributor": ["text", "image"],
     "moonshotai/Kimi-K2.5": ["text", "image"],
     "moonshotai/Kimi-K2.6": ["text", "image"],
     "moonshotai/Kimi-K2.7-Code": ["text", "image"],

package/dist/src/catalog/snapshot.js

@@ -6,6 +6,7 @@
 export const MODEL_SNAPSHOT = [
     { id: "claude-sonnet-5", name: "Claude Sonnet 5", contextLength: 1000000 },
     { id: "claude-sonnet-4-6", name: "Claude Sonnet 4.6", contextLength: 1000000 },
+    { id: "claude-fable-5-1", name: "Claude Fable 5.1", contextLength: 1000000 },
     { id: "claude-fable-5", name: "Claude Fable 5", contextLength: 1000000 },
     { id: "claude-opus-5", name: "Claude Opus 5", contextLength: 1000000 },
     { id: "claude-opus-4-8", name: "Claude Opus 4.8", contextLength: 1000000 },
@@ -21,6 +22,7 @@ export const MODEL_SNAPSHOT = [
     { id: "deepseek/deepseek-v4-pro", name: "DeepSeek V4 Pro (latest)", contextLength: 1000000 },
     { id: "deepseek/deepseek-v4-flash", name: "DeepSeek V4 Flash (latest)", contextLength: 1000000 },
     { id: "deepseek/deepseek-v4-flash-vision-exp", name: "DeepSeek V4 Flash Vision (exp)", contextLength: 1000000 },
+    { id: "deepseek/deepseek-v4-flash-fast", name: "DeepSeek V4 Flash Fast", contextLength: 1000000 },
     { id: "moonshotai/Kimi-K3", name: "Kimi K3", contextLength: 1000000 },
     { id: "moonshotai/Kimi-K2.7-Code", name: "Kimi K2.7 Code", contextLength: 256000 },
     { id: "moonshotai/Kimi-K2.7-Code-Highspeed", name: "Kimi K2.7 Code HighSpeed", contextLength: 262000 },
@@ -34,11 +36,10 @@ export const MODEL_SNAPSHOT = [
     { id: "zai-org/GLM-5", name: "GLM-5", contextLength: 200000 },
     { id: "MiniMaxAI/MiniMax-M3", name: "MiniMax M3", contextLength: 1000000 },
     { id: "MiniMaxAI/MiniMax-M2.7", name: "MiniMax M2.7", contextLength: 200000 },
-    { id: "minimax/minimax-m3-free", name: "MiniMax M3", contextLength: 1000000 },
-    { id: "minimax/minimax-m2.7-free", name: "MiniMax M2.7", contextLength: 197000 },
     { id: "MiniMaxAI/MiniMax-M2.5", name: "MiniMax M2.5", contextLength: 200000 },
     { id: "xiaomi/mimo-v2.5-pro", name: "MiMo V2.5 Pro", contextLength: 1000000 },
     { id: "xiaomi/mimo-v2.5", name: "MiMo V2.5", contextLength: 1000000 },
+    { id: "Qwen/Qwen3.8-Max-0902", name: "Qwen 3.8 Max 0902", contextLength: 1000000 },
     { id: "Qwen/Qwen3.8-Max", name: "Qwen 3.8 Max", contextLength: 1000000 },
     { id: "Qwen/Qwen3.8-27B", name: "Qwen 3.8 27B", contextLength: 262144 },
     { id: "Qwen/Qwen3.8-Flash", name: "Qwen 3.8 Flash", contextLength: 1000000 },
@@ -47,10 +48,12 @@ export const MODEL_SNAPSHOT = [
     { id: "Qwen/Qwen3.7-Flash", name: "Qwen 3.7 Flash", contextLength: 1000000 },
     { id: "Qwen/Qwen3.6-Max-Preview", name: "Qwen 3.6 Max Preview", contextLength: 200000 },
     { id: "Qwen/Qwen3.6-Plus", name: "Qwen 3.6 Plus", contextLength: 200000 },
+    { id: "meituan/LongCat-2.0:free", name: "LongCat 2.0", contextLength: 1048576 },
     { id: "stepfun/Step-3.7-Flash", name: "Step 3.7 Flash", contextLength: 256000 },
     { id: "stepfun/Step-3.5-Flash", name: "Step 3.5 Flash", contextLength: 1000000 },
     { id: "tencent/hy3-paid", name: "Tencent Hy3", contextLength: 262144 },
     { id: "tencent/hy4-preview", name: "Tencent Hy4 Preview", contextLength: 1048576 },
+    { id: "google/gemini-3.8-flash", name: "Gemini 3.8 Flash", contextLength: 1000000 },
     { id: "google/gemini-3.7-flash", name: "Gemini 3.7 Flash", contextLength: 1048576 },
     { id: "google/gemini-3.6-flash", name: "Gemini 3.6 Flash", contextLength: 1000000 },
     { id: "google/gemini-3.5-flash", name: "Gemini 3.5 Flash", contextLength: 1000000 },
@@ -64,6 +67,8 @@ export const MODEL_SNAPSHOT = [
     { id: "meta/muse-spark-1.1", name: "Muse Spark 1.1", contextLength: 1048576 },
     { id: "meta/muse-spark-1.2", name: "Muse Spark 1.2", contextLength: 1048576 },
     { id: "meta/muse-spark-1.2-contributor", name: "Muse Spark 1.2 Contributor", contextLength: 1048576 },
+    { id: "meta/muse-spark-1.3", name: "Muse Spark 1.3", contextLength: 1048576 },
+    { id: "meta/muse-spark-1.3-contributor", name: "Muse Spark 1.3 Contributor", contextLength: 1048576 },
     { id: "xai/grok-4.5", name: "Grok 4.5", contextLength: 500000 },
     { id: "xai/grok-4.6", name: "Grok 4.6", contextLength: 500000 },
 ];

package/dist/src/deals/catalog.d.ts

@@ -51,5 +51,5 @@ export interface ModelDeals {
 export declare const MODEL_DEALS: Readonly<Record<string, ModelDeals>>;
 export declare const PLAN_CATALOG: Readonly<Record<PlanId, PlanInfo>>;
 export declare const DEAL_SOURCE_URL = "https://commandcode.ai/docs/resources/pricing-limits";
-export declare const DEAL_LAST_REFRESHED = "2026-08-29";
+export declare const DEAL_LAST_REFRESHED = "2026-09-02";
 export declare const DEAL_PACKAGE_VERSION = "docs";

package/dist/src/deals/catalog.js

@@ -6,6 +6,7 @@
 // without network access at runtime. Regenerate with `npm run refresh:deals`.
 export const MODEL_DEALS = {
     "claude-fable-5": { tier: "premium", benchmark: { "intelligence": 62.1, "tokPerSec": 67.8 }, free: false, allowance: { "goat": 20, "pro": 20 } },
+    "claude-fable-5-1": { tier: "premium", free: false, allowance: { "goat": 20, "pro": 20 } },
     "claude-haiku-4-5-20251001": { tier: "premium", benchmark: { "intelligence": 24.1, "tokPerSec": 103.9 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "claude-opus-4-7": { tier: "premium", benchmark: { "intelligence": 55 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "claude-opus-4-8": { tier: "premium", benchmark: { "intelligence": 57.3 }, free: false, allowance: { "goat": 20, "pro": 20 } },
@@ -13,13 +14,15 @@ export const MODEL_DEALS = {
     "claude-sonnet-4-6": { tier: "premium", benchmark: { "intelligence": 48.4 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "claude-sonnet-5": { tier: "premium", benchmark: { "intelligence": 55.3, "tokPerSec": 89.6 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "deepseek/deepseek-v4-flash": { tier: "opensource", benchmark: { "intelligence": 51.8, "tokPerSec": 129 }, peakOffPeak: { "peak": { "input": 0.44, "output": 1.32, "cacheRead": 0.014, "cacheWrite": 0 }, "offPeak": { "input": 0.22, "output": 0.66, "cacheRead": 0.007, "cacheWrite": 0 }, "windows": "01–04 & 06–10 UTC" }, free: false, allowance: { "goat": 60, "pro": 70 } },
+    "deepseek/deepseek-v4-flash-fast": { tier: "opensource", free: false, allowance: { "goat": 20, "pro": 30 } },
     "deepseek/deepseek-v4-flash-vision-exp": { tier: "opensource", peakOffPeak: { "peak": { "input": 0.44, "output": 1.32, "cacheRead": 0.014, "cacheWrite": 0 }, "offPeak": { "input": 0.22, "output": 0.66, "cacheRead": 0.007, "cacheWrite": 0 }, "windows": "01–04 & 06–10 UTC" }, free: false, allowance: { "goat": 20, "pro": 30 } },
     "deepseek/deepseek-v4-pro": { tier: "opensource", benchmark: { "intelligence": 53.2, "tokPerSec": 61 }, peakOffPeak: { "peak": { "input": 1.32, "output": 3.96, "cacheRead": 0.044, "cacheWrite": 0 }, "offPeak": { "input": 0.66, "output": 1.98, "cacheRead": 0.022, "cacheWrite": 0 }, "windows": "01–04 & 06–10 UTC" }, free: false, allowance: { "goat": 20, "pro": 30 } },
     "google/gemini-3.1-flash-lite": { tier: "premium", benchmark: { "intelligence": 25.6 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "google/gemini-3.5-flash": { tier: "premium", benchmark: { "intelligence": 52 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "google/gemini-3.5-flash-lite": { tier: "premium", benchmark: { "intelligence": 37.4, "tokPerSec": 368.8 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "google/gemini-3.6-flash": { tier: "premium", benchmark: { "intelligence": 51.6, "tokPerSec": 197.3 }, free: false, allowance: { "goat": 20, "pro": 20 } },
-    "google/gemini-3.7-flash": { tier: "opensource", discount: { "pct": 50, "endsAt": "2026-12-31" }, was: { "input": 1.5, "output": 7.5, "cacheRead": 0.15 }, now: { "input": 0.75, "output": 3.75, "cacheRead": 0.075 }, benchmark: { "intelligence": 56, "tokPerSec": 365.9 }, free: false, allowance: { "goat": 20, "pro": 30 } },
+    "google/gemini-3.7-flash": { tier: "opensource", benchmark: { "intelligence": 56, "tokPerSec": 365.9 }, free: false, allowance: { "goat": 40, "pro": 50 } },
+    "google/gemini-3.8-flash": { tier: "premium", free: false, allowance: { "goat": 40, "pro": 50 } },
     "gpt-5.3-codex": { tier: "premium", benchmark: { "intelligence": 45.5, "tokPerSec": 121.4 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "gpt-5.4": { tier: "premium", benchmark: { "intelligence": 53.1 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "gpt-5.4-mini": { tier: "premium", benchmark: { "intelligence": 40.9 }, free: false, allowance: { "goat": 20, "pro": 20 } },
@@ -27,11 +30,12 @@ export const MODEL_DEALS = {
     "gpt-5.6-luna": { tier: "opensource", benchmark: { "intelligence": 52.3, "tokPerSec": 124.1 }, overContext: { "input": 0.4, "output": 1.8, "cacheRead": 0.04, "cacheWrite": 0.5 }, free: false, allowance: { "goat": 20, "pro": 30 } },
     "gpt-5.6-sol": { tier: "opensource", benchmark: { "intelligence": 60.9, "tokPerSec": 69.6 }, overContext: { "input": 10, "output": 45, "cacheRead": 1, "cacheWrite": 12.5 }, free: false, allowance: { "goat": 70, "pro": 80 } },
     "gpt-5.6-terra": { tier: "premium", benchmark: { "intelligence": 56.6, "tokPerSec": 112.8 }, overContext: { "input": 4, "output": 18, "cacheRead": 0.4, "cacheWrite": 5 }, free: false, allowance: { "goat": 20, "pro": 20 } },
+    "meituan/LongCat-2.0:free": { tier: "opensource", free: true },
     "meta/muse-spark-1.1": { tier: "premium", benchmark: { "intelligence": 53.2 }, free: false, allowance: { "goat": 20, "pro": 20 } },
     "meta/muse-spark-1.2": { tier: "opensource", benchmark: { "intelligence": 56.8 }, free: false, allowance: { "goat": 20, "pro": 30 } },
     "meta/muse-spark-1.2-contributor": { tier: "opensource", benchmark: { "intelligence": 56.8 }, free: false, allowance: { "goat": 20, "pro": 30 } },
-    "minimax/minimax-m2.7-free": { tier: "opensource", free: true, benchmark: { "intelligence": 38.9 } },
-    "minimax/minimax-m3-free": { tier: "opensource", free: true, benchmark: { "intelligence": 45.4 } },
+    "meta/muse-spark-1.3": { tier: "premium", free: false, allowance: { "goat": 20, "pro": 30 } },
+    "meta/muse-spark-1.3-contributor": { tier: "premium", free: false, allowance: { "goat": 20, "pro": 30 } },
     "MiniMaxAI/MiniMax-M2.5": { tier: "opensource", benchmark: { "intelligence": 34.5 }, free: false, allowance: { "goat": 20, "pro": 30 } },
     "MiniMaxAI/MiniMax-M2.7": { tier: "opensource", benchmark: { "intelligence": 38.9 }, free: false, allowance: { "goat": 20, "pro": 30 } },
     "MiniMaxAI/MiniMax-M3": { tier: "opensource", discount: { "pct": 50 }, was: { "input": 0.6, "output": 2.4, "cacheRead": 0.12 }, now: { "input": 0.3, "output": 1.2, "cacheRead": 0.06 }, benchmark: { "intelligence": 45.4, "tokPerSec": 111.3 }, free: false, allowance: { "goat": 47, "pro": 57 } },
@@ -50,6 +54,7 @@ export const MODEL_DEALS = {
     "Qwen/Qwen3.8-27B": { tier: "opensource", benchmark: { "intelligence": 52, "tokPerSec": 50.5 }, free: false, allowance: { "goat": 70, "pro": 80 } },
     "Qwen/Qwen3.8-Flash": { tier: "opensource", free: false, allowance: { "goat": 20, "pro": 30 } },
     "Qwen/Qwen3.8-Max": { tier: "opensource", benchmark: { "intelligence": 58.1, "tokPerSec": 23.6 }, free: false, allowance: { "goat": 20, "pro": 30 } },
+    "Qwen/Qwen3.8-Max-0902": { tier: "opensource", free: false, allowance: { "goat": 20, "pro": 30 } },
     "sakana/fugu-ultra": { tier: "premium", free: false, allowance: { "goat": 20, "pro": 20 } },
     "stepfun/Step-3.5-Flash": { tier: "opensource", benchmark: { "intelligence": 26.5 }, free: false, allowance: { "goat": 20, "pro": 30 } },
     "stepfun/Step-3.7-Flash": { tier: "opensource", benchmark: { "intelligence": 30.9, "tokPerSec": 94.1 }, free: false, allowance: { "goat": 20, "pro": 30 } },
@@ -78,5 +83,5 @@ export const PLAN_CATALOG = {
     provider: { price: 15, credits: 0, window5h: 0, windowWeek: 0, display: "Provider" },
 };
 export const DEAL_SOURCE_URL = "https://commandcode.ai/docs/resources/pricing-limits";
-export const DEAL_LAST_REFRESHED = "2026-08-29";
+export const DEAL_LAST_REFRESHED = "2026-09-02";
 export const DEAL_PACKAGE_VERSION = "docs";

package/dist/src/provider/reasoning.d.ts

@@ -4,10 +4,25 @@
 export declare const MODEL_EFFORTS: Readonly<Record<string, readonly CommandCodeReasoningEffort[]>>;
 export type PiThinkingLevel = "off" | "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
 type CommandCodeReasoningEffort = Exclude<PiThinkingLevel, "off">;
+/**
+ * Derives the reasoning-without-efforts set from a capability map and an
+ * efforts table: a model is reasoning-capable without explicit efforts
+ * when upstream's capability flag is true AND the facts carry no efforts
+ * entry for it. Efforts precedence holds by construction — an efforts
+ * model never appears in this set, so the "a model is classified exactly
+ * once" invariant is true by construction rather than asserted after
+ * regeneration (spec #108 / issue #111). Exported for the derivation
+ * checks; the runtime consumes the derived REASONING_MODELS below.
+ */
+export declare function deriveReasoningWithoutEfforts(capability: Readonly<Record<string, boolean>>, efforts: Readonly<Record<string, readonly string[]>>): ReadonlySet<string>;
 /**
  * Models Command Code advertises as reasoning-capable without exposing
  * explicit effort levels (Command Code chooses the reasoning depth). These
  * advertise `reasoning: true` in opencode but never generate `variants`.
+ *
+ * Derived from the generated classification module + the generated efforts
+ * facts — upstream data is the source of truth, so upstream classification
+ * changes land on the next catalog refresh with zero human edits.
  */
 export declare const REASONING_MODELS: ReadonlySet<string>;
 export interface ThinkingMetadata {
@@ -19,6 +34,11 @@ export interface ThinkingMetadata {
     };
 }
 export declare function thinkingMetadataForModel(modelId: string): ThinkingMetadata | undefined;
+/**
+ * A model is reasoning-capable when it has an explicit efforts entry OR
+ * upstream's capability flag is true. Derived entirely from the generated
+ * catalogs — no hand-maintained set remains.
+ */
 export declare function isReasoningModel(modelId: string): boolean;
 /**
  * opencode model `variants` for the Command Code model: one entry per supported

package/dist/src/provider/reasoning.js

@@ -1,8 +1,15 @@
 // src/provider/reasoning.ts — reasoning-effort metadata tables (PLAN #5 Part A,
-// port of pi's models.ts:67-150). Reasoning metadata now comes from the
-// generated catalog facts (`src/catalog/facts.ts`), with hand-maintained
-// classification sets (`REASONING_MODELS`) layered on top.
+// port of pi's models.ts:67-150). Reasoning metadata comes from the
+// generated catalog facts (`src/catalog/facts.ts`, the efforts table) and
+// the generated classification module (`src/catalog/classification.ts`,
+// the per-model capability flag the docs' RSC records carry). There is no
+// hand-maintained reasoning set: classification is **derived** (ADR-0006) —
+//   reasoning-without-efforts = capability flag true AND no efforts entry;
+//   isReasoningModel          = has an efforts entry OR capability flag true.
+// The only human seam is the classification override map baked into the
+// generated module at refresh time.
 import { MODEL_EFFORTS as GENERATED_MODEL_EFFORTS } from "../catalog/facts.js";
+import { MODEL_REASONING_CAPABILITY } from "../catalog/classification.js";
 /**
  * Models omitted here let Command Code choose their reasoning depth.
  */
@@ -14,39 +21,31 @@ import { MODEL_EFFORTS as GENERATED_MODEL_EFFORTS } from "../catalog/facts.js";
 // column is constrained to those levels; the release gate catches drift). Do
 // NOT "simplify" this back to a direct re-export — it breaks the build.
 export const MODEL_EFFORTS = GENERATED_MODEL_EFFORTS;
+/**
+ * Derives the reasoning-without-efforts set from a capability map and an
+ * efforts table: a model is reasoning-capable without explicit efforts
+ * when upstream's capability flag is true AND the facts carry no efforts
+ * entry for it. Efforts precedence holds by construction — an efforts
+ * model never appears in this set, so the "a model is classified exactly
+ * once" invariant is true by construction rather than asserted after
+ * regeneration (spec #108 / issue #111). Exported for the derivation
+ * checks; the runtime consumes the derived REASONING_MODELS below.
+ */
+export function deriveReasoningWithoutEfforts(capability, efforts) {
+    return new Set(Object.entries(capability)
+        .filter(([id, flag]) => flag === true && efforts[id] === undefined)
+        .map(([id]) => id));
+}
 /**
  * Models Command Code advertises as reasoning-capable without exposing
  * explicit effort levels (Command Code chooses the reasoning depth). These
  * advertise `reasoning: true` in opencode but never generate `variants`.
+ *
+ * Derived from the generated classification module + the generated efforts
+ * facts — upstream data is the source of truth, so upstream classification
+ * changes land on the next catalog refresh with zero human edits.
  */
-export const REASONING_MODELS = new Set([
-    "MiniMaxAI/MiniMax-M3",
-    "Qwen/Qwen3.6-Max-Preview",
-    "Qwen/Qwen3.6-Plus",
-    "Qwen/Qwen3.7-Flash",
-    "Qwen/Qwen3.7-Max",
-    "Qwen/Qwen3.7-Plus",
-    "meta/muse-spark-1.1",
-    "meta/muse-spark-1.2",
-    "meta/muse-spark-1.2-contributor",
-    "minimax/minimax-m3-free",
-    "moonshotai/Kimi-K2.7-Code",
-    "moonshotai/Kimi-K2.7-Code-Highspeed",
-    "moonshotai/Kimi-K3",
-    "nvidia/nemotron-3-ultra-550b-a55b",
-    "poolside/laguna-s-2.1-free",
-    "stepfun/Step-3.7-Flash",
-    "tencent/hy3-paid",
-    // tencent/hy4-preview is reasoning-capable with explicit efforts
-    // (models.md lists `low, medium, high` since command-code@1.38.0,
-    // 2026-08-28) — it belongs in MODEL_EFFORTS, not here. Keeping it here
-    // would leave a snapshot model classified in both REASONING_MODELS and
-    // MODEL_EFFORTS, which breaks the dedicated
-    // `tencent/hy4-preview is an efforts model` pin in
-    // tests/catalog-metadata.test.ts.
-    "thinkingmachines/inkling",
-    "thinkingmachines/inkling-small",
-]);
+export const REASONING_MODELS = deriveReasoningWithoutEfforts(MODEL_REASONING_CAPABILITY, GENERATED_MODEL_EFFORTS);
 const PI_THINKING_LEVELS = [
     "off",
     "minimal",
@@ -75,6 +74,11 @@ export function thinkingMetadataForModel(modelId) {
         },
     };
 }
+/**
+ * A model is reasoning-capable when it has an explicit efforts entry OR
+ * upstream's capability flag is true. Derived entirely from the generated
+ * catalogs — no hand-maintained set remains.
+ */
 export function isReasoningModel(modelId) {
     return MODEL_EFFORTS[modelId] !== undefined || REASONING_MODELS.has(modelId);
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-cmd-provider",
-  "version": "1.6.1",
+  "version": "1.6.3",
   "description": "Command Code provider + plugin for opencode",
   "type": "module",
   "main": "./dist/index.js",
@@ -20,9 +20,10 @@
     "typecheck": "tsc --noEmit",
     "refresh:snapshot": "node scripts/refresh-snapshot.mjs",
     "refresh:deals": "node scripts/refresh-deals.mjs",
-    "refresh": "npm run refresh:snapshot && npm run refresh:fixtures && npm run refresh:deals -- --fixtures",
+    "refresh:classification": "node scripts/refresh-classification.mjs",
+    "refresh": "npm run refresh:snapshot && npm run refresh:fixtures && npm run refresh:classification -- --fixtures && npm run refresh:deals -- --fixtures",
     "test": "npm run typecheck && npm run test:unit && npm run test:integration && npm run test:contract && npm run format:check",
-    "test:unit": "tsx tests/env.test.ts && tsx tests/auth-key.test.ts && tsx tests/converters.test.ts && tsx tests/stream.test.ts && tsx tests/provider-codecs.test.ts && tsx tests/provider-transport.test.ts && tsx tests/provider-parity.test.ts && tsx tests/provider-upgrade-fallback.test.ts && tsx tests/provider-zdr.test.ts && tsx tests/redact.test.ts && tsx tests/cost.test.ts && tsx tests/retry.test.ts && tsx tests/reasoning.test.ts && tsx tests/modalities.test.ts && tsx tests/snapshot.test.ts && tsx tests/refresh-snapshot.test.ts && tsx tests/oauth.test.ts && tsx tests/auth-mirror.test.ts && tsx tests/parse-facts.test.ts && tsx tests/parse-modalities.test.ts && tsx tests/catalog-metadata.test.ts && tsx tests/plugin-models.test.ts && tsx tests/vendor.test.ts && tsx tests/deals-enrichment.test.ts && tsx tests/deals-coverage.test.ts && tsx tests/plan-summary.test.ts && tsx tests/html-tables.test.ts && tsx tests/parse-docs.test.ts && tsx tests/parse-rsc.test.ts && tsx tests/mock-cc-rsc.test.ts && tsx tests/capture-rsc-fixtures.test.ts && tsx tests/refresh-deals.test.ts && tsx tests/refresh-deals-rsc.test.ts && tsx tests/tier-overrides.test.ts && tsx tests/tui-deals-panel.test.ts && tsx tests/diff-catalog.test.ts && tsx tests/catalog-refresh-extract.test.ts && tsx tests/catalog-refresh-push.test.ts && tsx tests/plugin-install.test.ts",
+    "test:unit": "tsx tests/env.test.ts && tsx tests/auth-key.test.ts && tsx tests/converters.test.ts && tsx tests/stream.test.ts && tsx tests/provider-codecs.test.ts && tsx tests/provider-transport.test.ts && tsx tests/provider-parity.test.ts && tsx tests/provider-upgrade-fallback.test.ts && tsx tests/provider-zdr.test.ts && tsx tests/redact.test.ts && tsx tests/cost.test.ts && tsx tests/retry.test.ts && tsx tests/reasoning.test.ts && tsx tests/modalities.test.ts && tsx tests/snapshot.test.ts && tsx tests/refresh-snapshot.test.ts && tsx tests/oauth.test.ts && tsx tests/auth-mirror.test.ts && tsx tests/parse-facts.test.ts && tsx tests/parse-modalities.test.ts && tsx tests/catalog-metadata.test.ts && tsx tests/plugin-models.test.ts && tsx tests/vendor.test.ts && tsx tests/deals-enrichment.test.ts && tsx tests/deals-coverage.test.ts && tsx tests/plan-summary.test.ts && tsx tests/html-tables.test.ts && tsx tests/parse-docs.test.ts && tsx tests/parse-rsc.test.ts && tsx tests/rsc-source.test.ts && tsx tests/classification-overrides.test.ts && tsx tests/refresh-classification.test.ts && tsx tests/classification.test.ts && tsx tests/mock-cc-rsc.test.ts && tsx tests/capture-rsc-fixtures.test.ts && tsx tests/refresh-deals.test.ts && tsx tests/refresh-deals-rsc.test.ts && tsx tests/tier-overrides.test.ts && tsx tests/tui-deals-panel.test.ts && tsx tests/diff-catalog.test.ts && tsx tests/catalog-refresh-extract.test.ts && tsx tests/catalog-refresh-drift.test.ts && tsx tests/catalog-refresh-push.test.ts && tsx tests/auto-release.test.ts && tsx tests/replay-verification.test.ts && tsx tests/plugin-install.test.ts",
     "test:integration": "tsx tests/integration-do-stream.test.ts && tsx tests/integration-do-generate.test.ts",
     "test:contract": "tsx tests/contract.test.ts",
     "test:e2e": "node tests/e2e-opencode.mjs",