chad-code 1.0.4__tar.gz → 1.0.6__tar.gz

{chad_code-1.0.4/src/chad_code.egg-info → chad_code-1.0.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: chad-code
-Version: 1.0.4
+Version: 1.0.6
 Summary: Local MLX-backed, Claude-Code-style coding agent (Apple Silicon, Ornith 35B/9B)
 License-Expression: MIT
 Project-URL: Repository, https://github.com/nathansutton/chad
@@ -21,7 +21,7 @@ Requires-Dist: mlx<0.33,>=0.32.0; sys_platform == "darwin"
 Requires-Dist: mlx-lm<0.32,>=0.31.3; sys_platform == "darwin"
 Requires-Dist: transformers<5.13,>=5.0
 Requires-Dist: jinja2>=3.1
-Requires-Dist: numpy<2,>=1.26
+Requires-Dist: numpy<3,>=1.26
 Requires-Dist: huggingface-hub>=1.22.0
 Requires-Dist: jedi>=0.19.2
 Requires-Dist: prompt_toolkit>=3.0.43
@@ -33,6 +33,8 @@ Provides-Extra: highlight
 Requires-Dist: pygments>=2.17; extra == "highlight"
 Provides-Extra: lsp
 Requires-Dist: serena-agent>=1.5; extra == "lsp"
+Provides-Extra: speech
+Requires-Dist: sounddevice>=0.5; extra == "speech"
 Dynamic: license-file
 
 # chad — a local, MLX-backed coding agent
@@ -58,7 +60,7 @@ uvx chad-code          # runs chad anywhere — the command is still `chad`
 uvx chad-code prove    # 2-min offline smoke test: 4 tiny fix-it tasks, verified, timed 🗿
 ```
 
-First run picks the right Ornith model for your RAM (9B under 32 GB, 35B at 32 GB+), asks,
+First run picks the right Ornith model for your RAM (9B under 24 GB, 35B at 24 GB+), asks,
 and downloads it once (~5 GB / ~13 GB, resumable) into the shared Hugging Face cache. While
 it downloads, `cd` into a project and think of a scoped first ask — *"fix the failing test
 in `tests/test_x.py`"* lands; *"improve my codebase"* flails.
@@ -81,10 +83,12 @@ instrument:
 | **Harness**     | open-ended, anything you can imagine            | plan. execute. nothing else.    |
 | **When wrong**  | reasons a way out                               | already shipped                 |
 
-![chad fixing a failing test end to end — reason, read, edit, rerun pytest, all on a local 35B](docs/demo.gif)
+![chad fixing a failing test end to end — reason, read, edit, run pytest, confirm green, all on a local 35B](docs/demo.gif)
 
-> Real session, unedited (cold model load cut): a local 35B reasons through the failure,
-> edits the file, reruns the tests, confirms green.
+> Real session, unedited (the silent prefill is cut): a local 35B finds the cent that
+> floor division loses, fixes it, and verifies itself — then your own `pytest`, in your
+> own shell. Recorded with `--yolo` so nothing pauses for a keypress; the default mode
+> stops and asks before every edit and every command.
 
 ## Frontier scores, laptop cost
 
@@ -139,16 +143,20 @@ uv run chad -c               # resume this directory's last conversation
 ```
 
 **The model.** chad picks one for you by RAM and downloads it once into the shared Hugging
-Face cache (`~/.cache/huggingface`, reused across every project). No picker, no flags —
-override with `CHAD_MODEL=<repo or local dir>` if you must.
+Face cache (`~/.cache/huggingface`, reused across every project). Override with
+`--model 9b` / `--model 35b`, or `--model <repo or local dir>` for anything else.
 
 | Your Mac | Model | Footprint |
 |---|---|---|
-| **≥ 32 GB** | [Ornith-1.0-35B `UD-Q2_K_XL`](https://huggingface.co/nathansutton/Ornith-1.0-35B-UD-Q2_K_XL-MLX) — 35B MoE, 2-bit experts | ~13 GB resident (~16 GB with KV) |
-| **16 / 18 / 24 GB** | [Ornith-1.0-9B `UD-Q4_K_XL`](https://huggingface.co/nathansutton/Ornith-1.0-9B-UD-Q4_K_XL-MLX) — 4-bit AWQ | ~5 GB |
-
-The 35B's working set needs headroom a 24 GB Mac doesn't have (it SIGKILLs mid-turn), so its
-floor is **32 GB**; 24 GB and below run the 9B. Quant names follow
+| **≥ 24 GB** | [Ornith-1.0-35B `UD-Q2_K_XL`](https://huggingface.co/nathansutton/Ornith-1.0-35B-UD-Q2_K_XL-MLX) — 35B MoE, 2-bit experts | ~13 GB resident (~16 GB with KV) |
+| **16 / 18 GB** | [Ornith-1.0-9B `UD-Q4_K_XL`](https://huggingface.co/nathansutton/Ornith-1.0-9B-UD-Q4_K_XL-MLX) — 4-bit AWQ | ~5 GB |
+
+The 35B's floor used to be 32 GB — its working set SIGKILLed a 24 GB Mac mid-turn. The
+fused attention kernel and the 8-bit-from-the-start KV cache it enables cut the per-token
+cache cost enough to give that headroom back, and the compaction trigger now sizes itself
+from the live Metal budget, so a tight box narrows its context window instead of dying.
+24 GB runs the 35B; 16/18 GB still get the 9B. If a 24 GB machine feels tight next to your
+other apps, `--model 9b` puts it back. Quant names follow
 [Unsloth's dynamic-quant convention](https://docs.unsloth.ai/) (`UD-…`).
 
 **Upgrading** — depends on how you installed: `uv tool upgrade chad-code`; `uvx --refresh
@@ -167,7 +175,10 @@ it chad uses the tree-sitter fallback automatically.
 `uv run chad` launches a full-screen terminal UI (built on prompt_toolkit):
 
 - **shift-tab cycles permission modes** — `normal` (confirm each bash/write/edit) →
-  `auto-accept edits` → `plan mode` (read-only: investigate + propose a numbered plan) → back.
+  `auto-accept edits` (edits land silently; **terminal commands still ask**) → `yolo`
+  (nothing asks) → `plan mode` (read-only: investigate + propose a numbered plan) → back.
+- **the approval prompt shows what you're approving** — the full command on its own lines
+  above the input, not a clipped single line. Approving blind isn't approving.
 - **type-ahead message queue** — keep typing while the agent works; messages run in order.
 - **ctrl-c interrupts the running turn** without killing the session.
 - **live status line** — model, mode, context %, a state glyph + verb, elapsed seconds, and
@@ -177,6 +188,14 @@ it chad uses the tree-sitter fallback automatically.
   `/model`, `/mode`, `/help`, `/exit`. Same set in the `--repl` line interface.
 - **`@file` / `@dir` mentions** and **`!command` shell passthrough** — pull a file into
   context inline, or run a shell command without invoking the model.
+- **voice mode, all local** — `/speech`, then ctrl-t to talk: Parakeet-on-MLX transcribes
+  into the input box for you to review before Enter sends it (esc discards a take);
+  replies are read aloud via macOS `say`. A pre-roll buffer means your first syllable
+  isn't clipped, and a personal word table (`~/.chad/speech_words.json`) teaches it your
+  identifiers — `{"pie test": "pytest"}`. Dictation cost is linear in take length, so a
+  long thought is fine; `/speech` off releases both the mic and the weights. Nothing
+  leaves the machine. Needs the `speech` extra (`uv sync --extra speech` — just a mic
+  library; no torch, no numba).
 
 **Usage.** `uv run chad --help` is the source of truth:
 
@@ -187,8 +206,17 @@ it chad uses the tree-sitter fallback automatically.
 | `--plan` | start in read-only plan mode (investigate + propose, edits blocked) |
 | `--yolo` | auto-approve bash/write/edit (skip confirm prompts) |
 | `--no-think` | skip Ornith's `<think>` blocks — faster on well-scoped work |
+| `--model` | `35b`, `9b`, `auto`, or any HF repo id / local model dir |
 | `--repl` | plain line REPL instead of the TUI |
 
+Plus three subcommands, each with its own `--help`:
+
+| Command | What it does |
+|---|---|
+| `chad serve` | serve this Mac's model to a container or the LAN ([Configuration](docs/configuration.md#serving-the-local-model-to-a-container-chad-serve)) |
+| `chad prove` | 2-minute offline smoke test: 4 tiny fix-it tasks, verified, timed |
+| `chad levers` | print the harness lever registry as JSON (ablation driver) |
+
 A headless task (positional, or piped with no TTY) auto-approves mutating tools; the model
 runs greedy (temp 0). Every conversation is persisted under `~/.chad/sessions/`, and every
 resume forks a new branch rather than overwriting — details in

{chad_code-1.0.4 → chad_code-1.0.6}/README.md

@@ -21,7 +21,7 @@ uvx chad-code          # runs chad anywhere — the command is still `chad`
 uvx chad-code prove    # 2-min offline smoke test: 4 tiny fix-it tasks, verified, timed 🗿
 ```
 
-First run picks the right Ornith model for your RAM (9B under 32 GB, 35B at 32 GB+), asks,
+First run picks the right Ornith model for your RAM (9B under 24 GB, 35B at 24 GB+), asks,
 and downloads it once (~5 GB / ~13 GB, resumable) into the shared Hugging Face cache. While
 it downloads, `cd` into a project and think of a scoped first ask — *"fix the failing test
 in `tests/test_x.py`"* lands; *"improve my codebase"* flails.
@@ -44,10 +44,12 @@ instrument:
 | **Harness**     | open-ended, anything you can imagine            | plan. execute. nothing else.    |
 | **When wrong**  | reasons a way out                               | already shipped                 |
 
-![chad fixing a failing test end to end — reason, read, edit, rerun pytest, all on a local 35B](docs/demo.gif)
+![chad fixing a failing test end to end — reason, read, edit, run pytest, confirm green, all on a local 35B](docs/demo.gif)
 
-> Real session, unedited (cold model load cut): a local 35B reasons through the failure,
-> edits the file, reruns the tests, confirms green.
+> Real session, unedited (the silent prefill is cut): a local 35B finds the cent that
+> floor division loses, fixes it, and verifies itself — then your own `pytest`, in your
+> own shell. Recorded with `--yolo` so nothing pauses for a keypress; the default mode
+> stops and asks before every edit and every command.
 
 ## Frontier scores, laptop cost
 
@@ -102,16 +104,20 @@ uv run chad -c               # resume this directory's last conversation
 ```
 
 **The model.** chad picks one for you by RAM and downloads it once into the shared Hugging
-Face cache (`~/.cache/huggingface`, reused across every project). No picker, no flags —
-override with `CHAD_MODEL=<repo or local dir>` if you must.
+Face cache (`~/.cache/huggingface`, reused across every project). Override with
+`--model 9b` / `--model 35b`, or `--model <repo or local dir>` for anything else.
 
 | Your Mac | Model | Footprint |
 |---|---|---|
-| **≥ 32 GB** | [Ornith-1.0-35B `UD-Q2_K_XL`](https://huggingface.co/nathansutton/Ornith-1.0-35B-UD-Q2_K_XL-MLX) — 35B MoE, 2-bit experts | ~13 GB resident (~16 GB with KV) |
-| **16 / 18 / 24 GB** | [Ornith-1.0-9B `UD-Q4_K_XL`](https://huggingface.co/nathansutton/Ornith-1.0-9B-UD-Q4_K_XL-MLX) — 4-bit AWQ | ~5 GB |
-
-The 35B's working set needs headroom a 24 GB Mac doesn't have (it SIGKILLs mid-turn), so its
-floor is **32 GB**; 24 GB and below run the 9B. Quant names follow
+| **≥ 24 GB** | [Ornith-1.0-35B `UD-Q2_K_XL`](https://huggingface.co/nathansutton/Ornith-1.0-35B-UD-Q2_K_XL-MLX) — 35B MoE, 2-bit experts | ~13 GB resident (~16 GB with KV) |
+| **16 / 18 GB** | [Ornith-1.0-9B `UD-Q4_K_XL`](https://huggingface.co/nathansutton/Ornith-1.0-9B-UD-Q4_K_XL-MLX) — 4-bit AWQ | ~5 GB |
+
+The 35B's floor used to be 32 GB — its working set SIGKILLed a 24 GB Mac mid-turn. The
+fused attention kernel and the 8-bit-from-the-start KV cache it enables cut the per-token
+cache cost enough to give that headroom back, and the compaction trigger now sizes itself
+from the live Metal budget, so a tight box narrows its context window instead of dying.
+24 GB runs the 35B; 16/18 GB still get the 9B. If a 24 GB machine feels tight next to your
+other apps, `--model 9b` puts it back. Quant names follow
 [Unsloth's dynamic-quant convention](https://docs.unsloth.ai/) (`UD-…`).
 
 **Upgrading** — depends on how you installed: `uv tool upgrade chad-code`; `uvx --refresh
@@ -130,7 +136,10 @@ it chad uses the tree-sitter fallback automatically.
 `uv run chad` launches a full-screen terminal UI (built on prompt_toolkit):
 
 - **shift-tab cycles permission modes** — `normal` (confirm each bash/write/edit) →
-  `auto-accept edits` → `plan mode` (read-only: investigate + propose a numbered plan) → back.
+  `auto-accept edits` (edits land silently; **terminal commands still ask**) → `yolo`
+  (nothing asks) → `plan mode` (read-only: investigate + propose a numbered plan) → back.
+- **the approval prompt shows what you're approving** — the full command on its own lines
+  above the input, not a clipped single line. Approving blind isn't approving.
 - **type-ahead message queue** — keep typing while the agent works; messages run in order.
 - **ctrl-c interrupts the running turn** without killing the session.
 - **live status line** — model, mode, context %, a state glyph + verb, elapsed seconds, and
@@ -140,6 +149,14 @@ it chad uses the tree-sitter fallback automatically.
   `/model`, `/mode`, `/help`, `/exit`. Same set in the `--repl` line interface.
 - **`@file` / `@dir` mentions** and **`!command` shell passthrough** — pull a file into
   context inline, or run a shell command without invoking the model.
+- **voice mode, all local** — `/speech`, then ctrl-t to talk: Parakeet-on-MLX transcribes
+  into the input box for you to review before Enter sends it (esc discards a take);
+  replies are read aloud via macOS `say`. A pre-roll buffer means your first syllable
+  isn't clipped, and a personal word table (`~/.chad/speech_words.json`) teaches it your
+  identifiers — `{"pie test": "pytest"}`. Dictation cost is linear in take length, so a
+  long thought is fine; `/speech` off releases both the mic and the weights. Nothing
+  leaves the machine. Needs the `speech` extra (`uv sync --extra speech` — just a mic
+  library; no torch, no numba).
 
 **Usage.** `uv run chad --help` is the source of truth:
 
@@ -150,8 +167,17 @@ it chad uses the tree-sitter fallback automatically.
 | `--plan` | start in read-only plan mode (investigate + propose, edits blocked) |
 | `--yolo` | auto-approve bash/write/edit (skip confirm prompts) |
 | `--no-think` | skip Ornith's `<think>` blocks — faster on well-scoped work |
+| `--model` | `35b`, `9b`, `auto`, or any HF repo id / local model dir |
 | `--repl` | plain line REPL instead of the TUI |
 
+Plus three subcommands, each with its own `--help`:
+
+| Command | What it does |
+|---|---|
+| `chad serve` | serve this Mac's model to a container or the LAN ([Configuration](docs/configuration.md#serving-the-local-model-to-a-container-chad-serve)) |
+| `chad prove` | 2-minute offline smoke test: 4 tiny fix-it tasks, verified, timed |
+| `chad levers` | print the harness lever registry as JSON (ablation driver) |
+
 A headless task (positional, or piped with no TTY) auto-approves mutating tools; the model
 runs greedy (temp 0). Every conversation is persisted under `~/.chad/sessions/`, and every
 resume forks a new branch rather than overwriting — details in

{chad_code-1.0.4 → chad_code-1.0.6}/pyproject.toml

@@ -4,7 +4,7 @@
 # import name, and command name are independent. `uvx chad-code` runs the alias
 # script added under [project.scripts].
 name = "chad-code"
-version = "1.0.4"
+version = "1.0.6"
 description = "Local MLX-backed, Claude-Code-style coding agent (Apple Silicon, Ornith 35B/9B)"
 readme = "README.md"
 license = "MIT"
@@ -49,7 +49,10 @@ dependencies = [
     # absent in a clean install (e.g. chad inside a Linux benchmark container), where its
     # absence crashes the first turn. Declare it directly.
     "jinja2>=3.1",
-    "numpy>=1.26,<2",
+    # <3: the old <2 cap was 1.0.0-era caution, not a known break. numpy 2.4.6
+    # verified end-to-end (full gate + live Parakeet-on-MLX transcription +
+    # `chad prove` 4/4); staying open keeps 2.x-floored ASR deps adoptable.
+    "numpy>=1.26,<3",
     "huggingface-hub>=1.22.0", # first-run model download (cli.py/engine.py); imported directly, not just transitively
     "jedi>=0.19.2", # in-process symbolic code intelligence (read/search/edit)
     "prompt_toolkit>=3.0.43",
@@ -78,6 +81,16 @@ highlight = ["pygments>=2.17"]
 # heavyweight tree; absent, every caller falls back to the tree-sitter backend.
 # Install with `uv sync --extra lsp`.
 lsp = ["serena-agent>=1.5"]
+# Voice mode (/speech in the TUI): push-to-talk dictation via chad's vendored
+# Parakeet-on-MLX (src/chad/parakeet — the PyPI parakeet-mlx drags librosa's
+# numba/scipy tree for one mel-filterbank call) + spoken replies via macOS
+# `say`. All on-device, nothing leaves the machine. mlx/numpy/huggingface-hub
+# are already core deps, so the extra adds only the mic (sounddevice/PortAudio).
+# Optional and import-guarded in speech.py: absent, /speech prints the install
+# hint and everything else is untouched. Install with `uv sync --extra speech`.
+speech = [
+    "sounddevice>=0.5",
+]
 
 # The one entrypoint: `uv run chad`. (`chad-bench` reproduces docs/benchmarks.md.)
 # `chad-code` is an alias so `uvx chad-code` (which runs the console script matching
@@ -107,6 +120,10 @@ build-backend = "setuptools.build_meta"
 [tool.setuptools.packages.find]
 where = ["src"]
 
+# The vendored Parakeet code is Apache-2.0; its license text must travel with
+# the code, wheel included.
+[tool.setuptools.package-data]
+"chad.parakeet" = ["LICENSE"]
 [tool.pytest.ini_options]
 testpaths = ["tests"]
 # No "-q" here: CI/docs/agents all invoke `pytest -q` explicitly, and stacking the two
@@ -127,6 +144,12 @@ src = ["src", "tests"]
 select = ["F", "E9", "I"]
 ignore = ["E741"]  # ambiguous var names (l/I) appear in existing math code
 
+[tool.ruff.lint.per-file-ignores]
+# Vendored upstream (see chad/parakeet/__init__.py), kept byte-close for
+# future re-syncs — not held to chad's own lint standard. (F401: imports left
+# in place for upstream fidelity after the transcribe(path) cut.)
+"src/chad/parakeet/*" = ["F401", "F821", "F841", "E741"]
+
 [tool.mypy]
 python_version = "3.11"
 files = ["src/chad"]
@@ -142,6 +165,12 @@ disable_error_code = ["var-annotated"]
 # Tighten (disallow_untyped_defs, re-enable var-annotated, add B/UP/SIM in ruff) in a
 # follow-up once this baseline holds.
 
+[[tool.mypy.overrides]]
+# Vendored upstream inference code (see chad/parakeet/__init__.py) is not held
+# to chad's gate so re-syncs stay a plain re-copy.
+module = "chad.parakeet.*"
+ignore_errors = true
+
 [[tool.mypy.overrides]]
 # mlx_fastpath patches mlx_lm classes that only exist on macOS: with mlx installed
 # (dev Macs, the macos CI jobs) its `type: ignore[method-assign]` comments are load-

chad_code-1.0.6/src/chad/__init__.py

@@ -0,0 +1,14 @@
+"""chad — a local MLX-backed, Claude-Code-style coding agent.
+
+A flat collection of cooperating modules behind one console script (``chad``):
+the inference engine, the tool layer, the agent loop, and the terminal UI.
+"""
+
+__version__ = "1.0.6"
+
+# chad sets no MLX_* runtime vars. MLX_METAL_FAST_SYNCH, MLX_MAX_OPS_PER_BUFFER
+# and MLX_MAX_MB_PER_BUFFER were each measured end-to-end on the 35B and every
+# setting was slower than mlx's own defaults (1-4% at both 8k and 32k), so the
+# mechanism that used to apply them at package import is gone. Any future entry
+# has to be set before mlx's backend initializes — package import is the last
+# hook early enough, since chad imports mlx lazily inside functions.