simplicio-cli 0.5.17__tar.gz → 0.6.0__tar.gz

{simplicio_cli-0.5.17 → simplicio_cli-0.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: simplicio-cli
-Version: 0.5.17
+Version: 0.6.0
 Summary: Portable task-to-code pipeline that works with any LLM. Turn a one-line task into a verified code change — diff + test + verify loop. +55 pts on a 156-check benchmark, 21% faster, ~same tokens.
 Author-email: Wesley Simplicio <wesleybob4@gmail.com>
 License: MIT
@@ -26,21 +26,21 @@ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: sentence-transformers>=5.1.2
-Requires-Dist: numpy>=2.0.2
-Requires-Dist: anthropic>=0.105.2
-Requires-Dist: openai>=2.38.0
-Requires-Dist: simplicio-mapper>=0.7.3
-Requires-Dist: simplicio-prompt>=1.13.3
+Requires-Dist: sentence-transformers>=5.6.0
+Requires-Dist: numpy>=2.5.0
+Requires-Dist: anthropic>=0.112.0
+Requires-Dist: openai>=2.44.0
+Requires-Dist: simplicio-mapper>=0.9.0
+Requires-Dist: simplicio-prompt>=1.14.1
 Requires-Dist: httpx>=0.28.1
-Requires-Dist: orjson>=3.11.5
+Requires-Dist: orjson>=3.11.9
 Requires-Dist: diskcache>=5.6.3
 Requires-Dist: libcst>=1.8.6
 Provides-Extra: bench
-Requires-Dist: fpdf2>=2.8.4; extra == "bench"
+Requires-Dist: fpdf2>=2.8.7; extra == "bench"
 Provides-Extra: local
-Requires-Dist: llama-cpp-python>=0.3.23; extra == "local"
-Requires-Dist: huggingface-hub>=1.8.0; extra == "local"
+Requires-Dist: llama-cpp-python>=0.3.32; extra == "local"
+Requires-Dist: huggingface-hub>=1.21.0; extra == "local"
 Dynamic: license-file
 
 <h1 align="center">simplicio-cli</h1>
@@ -81,8 +81,8 @@ The new first screen is the doorway; the restored guide below is the workshop. T
 
 ```bash
 pip install -U simplicio-cli
-simplicio detect "hide the Delete button for non-admins"
-simplicio task "hide the Delete button for non-admins"
+simplicio-py detect "hide the Delete button for non-admins"
+simplicio-py task "hide the Delete button for non-admins"
 ```
 
 ## What it does
@@ -482,18 +482,18 @@ simplicio-cli has **three distinct entry points**. Same engine, three front door
 | You have | Path | LLM call goes through | Need API key? |
 |---|---|---|---|
 | **Claude Code** (Pro / Max / Team / API) | Skill + hook auto-installed in `~/.claude/` | Claude Code itself, using your logged-in session | **No** |
-| **Claude Code OAuth or Codex CLI / ChatGPT Plus** | `simplicio task` with `SIMPLICIO_MODEL=claude-cli/<m>` or `codex-cli/<m>` | Shell-out to `claude -p` / `codex exec` (subprocess uses your existing login) | **No** |
-| **API key** for any provider (Anthropic, OpenAI, OpenRouter, GLM, DeepSeek, Ollama…) | `simplicio task` standalone CLI | The provider SDK directly | **Yes** — set `SIMPLICIO_API_KEY` |
+| **Claude Code OAuth or Codex CLI / ChatGPT Plus** | `simplicio-py task` with `SIMPLICIO_MODEL=claude-cli/<m>` or `codex-cli/<m>` | Shell-out to `claude -p` / `codex exec` (subprocess uses your existing login) | **No** |
+| **API key** for any provider (Anthropic, OpenAI, OpenRouter, GLM, DeepSeek, Ollama…) | `simplicio-py task` standalone CLI | The provider SDK directly | **Yes** — set `SIMPLICIO_API_KEY` |
 
-**Most users land on Path 1.** `pip install simplicio-cli` puts the binary on PATH; the first invocation auto-installs the skill + hook in `~/.claude/` (idempotent, opt-out via `SIMPLICIO_SKIP_AUTO_INIT=1`). From that moment, every code-edit prompt you type **inside Claude Code** is silently routed through simplicio's 6-layer contract — no extra config, no key, no cost beyond your existing Claude subscription.
+**Most users land on Path 1.** `pip install simplicio-cli` puts `simplicio-py` on PATH; the first invocation auto-installs the skill + hook in `~/.claude/` (idempotent, opt-out via `SIMPLICIO_SKIP_AUTO_INIT=1`). From that moment, every code-edit prompt you type **inside Claude Code** is silently routed through simplicio's 6-layer contract — no extra config, no key, no cost beyond your existing Claude subscription.
 
-**Path 2 — subscription shell-out (zero key).** If you have a Claude Pro/Max session (`claude login`) or a ChatGPT Plus + Codex CLI session (`codex login`) and want to drive simplicio from CI, scripts, or any context **outside** Claude Code, set `SIMPLICIO_MODEL=claude-cli/<model>` or `codex-cli/<model>`. simplicio spawns the CLI as a subprocess; the call rides your existing OAuth session — no API key required. A recursion guard (`SIMPLICIO_HOOK_GUARD=1`) is injected so the inner CLI does not re-fire simplicio's own hook.
+**Path 2 — subscription shell-out (zero key).** If you have a Claude Pro/Max session (`claude login`) or a ChatGPT Plus + Codex CLI session (`codex login`) and want to drive simplicio from CI, scripts, or any context **outside** Claude Code, set `SIMPLICIO_MODEL=claude-cli/<model>` or `codex-cli/<model>`. `simplicio-py` spawns the CLI as a subprocess; the call rides your existing OAuth session — no API key required. A recursion guard (`SIMPLICIO_HOOK_GUARD=1`) is injected so the inner CLI does not re-fire the hook.
 
-**Path 3 is for environments without any logged-in CLI** — a remote server, a build runner, a notebook, a different LLM provider. You bring an API key (Anthropic, OpenRouter, OpenAI, GLM, DeepSeek, Ollama…), simplicio calls the provider directly.
+**Path 3 is for environments without any logged-in CLI** — a remote server, a build runner, a notebook, a different LLM provider. You bring an API key (Anthropic, OpenRouter, OpenAI, GLM, DeepSeek, Ollama…), `simplicio-py` calls the provider directly.
 
 #### Path 1 example — inside Claude Code
 
-After `pip install simplicio-cli && simplicio smoke` (which triggers auto-bootstrap), just type your task in Claude Code:
+After `pip install simplicio-cli && simplicio-py smoke` (which triggers auto-bootstrap), just type your task in Claude Code:
 
 ```
 hide the Delete button for non-admins in src/app/screen/screen.component.html
@@ -511,27 +511,27 @@ piggybacks on that login — no extra bill, no key to manage.
 export SIMPLICIO_MODEL=claude-cli/sonnet     # or claude-cli/opus, claude-cli/default
 unset  SIMPLICIO_API_KEY                     # explicitly: no key needed
 
-simplicio task "hide Delete button for non-admins" --stack angular \
+simplicio-py task "hide Delete button for non-admins" --stack angular \
   --target src/app/screen/screen.component.html
 
 # Option B — Codex CLI subscription (run `codex login` once)
 export SIMPLICIO_MODEL=codex-cli/gpt-5       # or codex-cli/default
-simplicio task "..." --stack angular --target ...
+simplicio-py task "..." --stack angular --target ...
 ```
 
-How it works: simplicio shells out to `claude -p "<prompt>"` (or `codex exec "<prompt>"`) as a subprocess, captures stdout, runs the test loop. The inner CLI authenticates via your existing OAuth session in `~/.claude/` or `~/.codex/`. simplicio sets `SIMPLICIO_HOOK_GUARD=1` in the subprocess env so the inner Claude Code session does **not** re-fire simplicio's own UserPromptSubmit hook (no infinite recursion).
+How it works: `simplicio-py` shells out to `claude -p "<prompt>"` (or `codex exec "<prompt>"`) as a subprocess, captures stdout, runs the test loop. The inner CLI authenticates via your existing OAuth session in `~/.claude/` or `~/.codex/`. `simplicio-py` sets `SIMPLICIO_HOOK_GUARD=1` in the subprocess env so the inner Claude Code session does **not** re-fire its own UserPromptSubmit hook (no infinite recursion).
 
-For orchestrators such as SendSprint, `simplicio task` also has a structured
+For orchestrators such as SendSprint, `simplicio-py task` also has a structured
 contract:
 
 ```bash
-simplicio task "hide Delete button for non-admins" \
+simplicio-py task "hide Delete button for non-admins" \
   --stack angular \
   --target src/app/screen/screen.component.html \
   --dry-run-task \
   --json
 
-simplicio task "front-only task" \
+simplicio-py task "front-only task" \
   --stack angular \
   --target src/app/screen/screen.component.html \
   --bound-paths "src/app/**" \
@@ -551,8 +551,8 @@ export SIMPLICIO_API_KEY=sk-or-v1-…                      # OpenRouter key
 export SIMPLICIO_MODEL=anthropic/claude-opus-4
 export SIMPLICIO_BASE_URL=https://openrouter.ai/api/v1
 
-simplicio index --stack angular                           # one-time, builds embedding cache
-simplicio task "hide Delete button for non-admins" \
+simplicio-py index --stack angular                           # one-time, builds embedding cache
+simplicio-py task "hide Delete button for non-admins" \
   --stack angular \
   --target src/app/screen/screen.component.html \
   --criteria "- no admin perm: button absent from DOM
@@ -567,20 +567,20 @@ Provider-agnostic — see [Configure](#configure--any-llm-nothing-hardcoded) for
 
 #### Path 1 deep-dive — auto-activation in Claude Code
 
-`pip install` puts `simplicio` on your PATH. To make Claude Code
+`pip install` puts `simplicio-py` on your PATH. To make Claude Code
 **automatically** route code-edit tasks through simplicio, a skill + hook
 need to land in `~/.claude/`.
 
-**Zero-step path (recommended).** The first time you run *any* `simplicio`
+**Zero-step path (recommended).** The first time you run *any* `simplicio-py`
 command after install, if Claude Code is present (`~/.claude/` exists) and
-the hook is missing, simplicio installs both for you and prints one stderr
+the hook is missing, `simplicio-py` installs both for you and prints one stderr
 line. PEP 517 wheels can't execute code on `pip install`, so this is the
 closest equivalent that works on every machine.
 
 ```bash
 pip install simplicio-cli
-simplicio smoke         # ← first call also installs skill + hook (idempotent)
-# stderr: "simplicio: auto-activation installed in Claude Code …"
+simplicio-py smoke         # ← first call also installs skill + hook (idempotent)
+# stderr: "simplicio-py: auto-activation installed in Claude Code …"
 ```
 
 Opt out before the first call:
@@ -592,9 +592,9 @@ export SIMPLICIO_SKIP_AUTO_INIT=1
 **Explicit path.** Same effect, no auto-magic:
 
 ```bash
-simplicio init                 # idempotent
-simplicio init --dry-run       # preview only
-simplicio init --claude-home <path>   # override target dir
+simplicio-py init                 # idempotent
+simplicio-py init --dry-run       # preview only
+simplicio-py init --claude-home <path>   # override target dir
 ```
 
 Either way, two files land in `~/.claude/`:
@@ -602,7 +602,7 @@ Either way, two files land in `~/.claude/`:
 | File | Purpose |
 |---|---|
 | `~/.claude/skills/simplicio-cli/SKILL.md` | Skill the agent matches by description when your prompt looks like a code edit |
-| `~/.claude/hooks/simplicio-userpromptsubmit.sh` + entry in `~/.claude/settings.json` | UserPromptSubmit hook that runs `simplicio detect` on every prompt and injects a hint when the heuristic catches a code-edit task the skill could miss |
+| `~/.claude/hooks/simplicio-userpromptsubmit.sh` + entry in `~/.claude/settings.json` | UserPromptSubmit hook that runs `simplicio-py detect` on every prompt and injects a hint when the heuristic catches a code-edit task the skill could miss |
 
 A backup of your previous `settings.json` is written to `settings.json.bak`
 before any merge.
@@ -613,13 +613,13 @@ After install, every prompt you type in Claude Code flows through two layers:
 
 1. **Skill layer (semantic).** Claude reads the SKILL.md description. When
    your prompt looks like a programming task ("add X to Y.tsx", "fix the auth
-   bug in middleware.py"), Claude considers using `simplicio task` instead of
+   bug in middleware.py"), Claude considers using `simplicio-py task` instead of
    writing code directly.
-2. **Hook layer (deterministic).** Every prompt fires `simplicio detect` via
+2. **Hook layer (deterministic).** Every prompt fires `simplicio-py detect` via
    the UserPromptSubmit hook. The classifier scores the prompt (verbs + file
    extensions + code nouns − read-only cues). Score ≥ 3 → it emits a
    `[SIMPLICIO_PROMPT_HINT]` block on stderr. Claude sees the hint alongside
-   your prompt — a hard nudge toward `simplicio task <prompt> <repo>`.
+   your prompt — a hard nudge toward `simplicio-py task <prompt> <repo>`.
 
 The layers are complementary. Skill = "Claude *might* pick simplicio". Hook
 = "Claude *sees* the hint regardless".
@@ -635,11 +635,11 @@ user prompt. UserPromptSubmit is the right pre-hook for routing decisions.
 
 | Goal | How |
 |---|---|
-| Block the auto-bootstrap | `export SIMPLICIO_SKIP_AUTO_INIT=1` before the first `simplicio` call |
+| Block the auto-bootstrap | `export SIMPLICIO_SKIP_AUTO_INIT=1` before the first `simplicio-py` call |
 | Disable hook permanently | Delete `~/.claude/hooks/simplicio-userpromptsubmit.sh` and its entry in `~/.claude/settings.json` |
-| Re-install / repair | `simplicio init` (idempotent — won't double-write) |
-| Preview without writing | `simplicio init --dry-run` |
-| Skill-only (no hook) | Copy `.skills/simplicio-cli/SKILL.md` to `~/.claude/skills/simplicio-cli/SKILL.md` manually, skip `simplicio init` |
+| Re-install / repair | `simplicio-py init` (idempotent — won't double-write) |
+| Preview without writing | `simplicio-py init --dry-run` |
+| Skill-only (no hook) | Copy `.skills/simplicio-cli/SKILL.md` to `~/.claude/skills/simplicio-cli/SKILL.md` manually, skip `simplicio-py init` |
 
 ---
 
@@ -655,7 +655,7 @@ user prompt. UserPromptSubmit is the right pre-hook for routing decisions.
 | GLM (z.ai) | `glm-4.6` | `https://api.z.ai/api/paas/v4` |
 | DeepSeek | `deepseek-chat` | `https://api.deepseek.com` |
 | OpenAI | `gpt-4.1` | `https://api.openai.com/v1` |
-| Local (llama.cpp) | `local-llama/default` | *(leave unset)* |
+| Local (llama.cpp) | `openbmb/minicpm5:latest` | *(leave unset)* |
 | Anthropic native | `claude-opus-4-7` | *(leave unset)* |
 
 If `SIMPLICIO_BASE_URL` is unset and the key is `ANTHROPIC_API_KEY`, it uses the
@@ -663,7 +663,7 @@ native Anthropic SDK. Otherwise it uses an OpenAI-compatible client pointed at
 your `base_url` — so **any** OpenAI-like provider works without code changes.
 
 ```bash
-simplicio smoke      # prints provider config + one test call
+simplicio-py smoke      # prints provider config + one test call
 ```
 
 #### Path 4 — local llama.cpp GGUF default
@@ -671,14 +671,14 @@ simplicio smoke      # prints provider config + one test call
 When **no provider is configured** (`SIMPLICIO_MODEL` and
 `SIMPLICIO_BASE_URL` both unset), simplicio runs the in-process
 [`llama-cpp-python`](https://github.com/abetlen/llama-cpp-python) backend with
-`local-llama/default`, currently
-`bartowski/Qwen_Qwen3.5-2B-GGUF::Qwen_Qwen3.5-2B-Q6_K.gguf`.
+`openbmb/minicpm5:latest`, backed by
+`openbmb/MiniCPM5-1B-GGUF::MiniCPM5-1B-Q4_K_M.gguf`.
 
 ```bash
 pip install 'simplicio-cli[local]'          # pulls llama-cpp-python + huggingface-hub
-simplicio doctor --install                  # downloads/validates the default GGUF
+simplicio-py doctor --install                  # downloads/validates the default GGUF
 
-simplicio task "add input validation to createUser" \
+simplicio-py task "add input validation to createUser" \
   --target src/users.ts --local              # forces local llama.cpp
 
 # the GGUF is fetched once from the Hugging Face Hub, then reused
@@ -687,17 +687,24 @@ simplicio task "add input validation to createUser" \
 Explicit routes (override the default model/weights):
 
 ```bash
-SIMPLICIO_MODEL=local-llama/default                                  # Qwen_Qwen3.5-2B-Q6_K.gguf default
-SIMPLICIO_MODEL=local-llama/bartowski/Qwen_Qwen3.5-2B-GGUF::Qwen_Qwen3.5-2B-Q6_K.gguf
+SIMPLICIO_MODEL=openbmb/minicpm5:latest                              # MiniCPM5-1B-Q4_K_M.gguf default
+SIMPLICIO_MODEL=local-llama/default                                  # backward-compatible alias
+SIMPLICIO_MODEL=local-llama/openbmb/MiniCPM5-1B-GGUF::MiniCPM5-1B-Q4_K_M.gguf
 SIMPLICIO_MODEL=local-llama//models/my-model.gguf                    # direct local path
 SIMPLICIO_LOCAL_MODEL_PATH=/models/my-model.gguf                     # always wins
 ```
 
 Tuning knobs (all optional): `SIMPLICIO_LOCAL_CTX` (context window, default
-`8192`), `SIMPLICIO_LOCAL_THREADS`, `SIMPLICIO_LOCAL_GPU_LAYERS` (offload to GPU,
-default `0`), `SIMPLICIO_LOCAL_MAX_TOKENS` (generation cap),
-`SIMPLICIO_LOCAL_TEMP` (default `0.1`), `SIMPLICIO_LOCAL_MODEL_REPO` /
-`SIMPLICIO_LOCAL_MODEL_FILE`.
+`2048`, clamped by `SIMPLICIO_LOCAL_CTX_MAX`, default `4096`),
+`SIMPLICIO_LOCAL_THREADS` (default and cap `4` via
+`SIMPLICIO_LOCAL_THREADS_MAX`), `SIMPLICIO_LOCAL_GPU_LAYERS` (offload to GPU,
+default `0`), `SIMPLICIO_LOCAL_BATCH` (default/cap `128`),
+`SIMPLICIO_LOCAL_UBATCH` (default/cap `32`), `SIMPLICIO_LOCAL_MAX_TOKENS`
+(generation cap, default `512`, clamped by `SIMPLICIO_LOCAL_MAX_TOKENS_CAP`,
+default `2048`), `SIMPLICIO_LOCAL_TEMP` (default `0.1`),
+`SIMPLICIO_LOCAL_MODEL_REPO` / `SIMPLICIO_LOCAL_MODEL_FILE`. The runtime keeps
+`mmap` enabled and `mlock` disabled so `llama.cpp` does not accidentally
+over-allocate RAM.
 
 #### The pipeline (both paths)
 
@@ -728,10 +735,10 @@ prompt. No key needed.
 **"I want to run it in CI / a script / outside Claude Code."** Path 2. Get an
 API key from any of the providers above (OpenRouter is the cheapest way to
 try multiple models behind one key), set `SIMPLICIO_API_KEY` +
-`SIMPLICIO_MODEL` + optional `SIMPLICIO_BASE_URL`, run `simplicio task ...`.
+`SIMPLICIO_MODEL` + optional `SIMPLICIO_BASE_URL`, run `simplicio-py task ...`.
 
 **"How do I load `.env.local` safely before running a local API?"** Use
-`eval "$(simplicio env-export .env.local)"` instead of `source .env.local`.
+`eval "$(simplicio-py env-export .env.local)"` instead of `source .env.local`.
 This preserves values with semicolons, such as PostgreSQL connection strings,
 without executing the dotenv file as shell code.
 
@@ -743,7 +750,7 @@ tracked, not shipped.
 
 **"Will Claude Code use simplicio for *every* prompt now?"** No. The skill
 only triggers on prompts that look like code edits (the description is
-specific). The hook fires `simplicio detect` on every prompt but only emits
+specific). The hook fires `simplicio-py detect` on every prompt but only emits
 a hint when the deterministic classifier scores ≥ 3 (verbs + file extensions
 + code nouns − read-only cues). "What does this function do?" gets no
 nudge. "Add a delete confirmation to UserList.tsx" does.
@@ -783,7 +790,7 @@ block, TEST block, contract-state words. Full numbers in [`bench/results.md`](be
 #### Full harness (your real project, your real tests)
 
 ```bash
-simplicio bench --cases bench/cases.json --stack angular
+simplicio-py bench --cases bench/cases.json --stack angular
 ```
 
 Runs each case two ways and runs **your real test command** (e.g. `ng test
@@ -897,7 +904,7 @@ bench/
   run_offline.py  # stdlib-only multi-model benchmark
   cases.json      # your benchmark tasks
   cases_offline.json
-  results.md      # filled by `simplicio bench` / `run_offline.py`
+  results.md      # filled by `simplicio-py bench` / `run_offline.py`
   charts/         # SVG: overall, delta, by_case, by_stack
 ```
 

{simplicio_cli-0.5.17 → simplicio_cli-0.6.0}/README.md

@@ -36,8 +36,8 @@ The new first screen is the doorway; the restored guide below is the workshop. T
 
 ```bash
 pip install -U simplicio-cli
-simplicio detect "hide the Delete button for non-admins"
-simplicio task "hide the Delete button for non-admins"
+simplicio-py detect "hide the Delete button for non-admins"
+simplicio-py task "hide the Delete button for non-admins"
 ```
 
 ## What it does
@@ -437,18 +437,18 @@ simplicio-cli has **three distinct entry points**. Same engine, three front door
 | You have | Path | LLM call goes through | Need API key? |
 |---|---|---|---|
 | **Claude Code** (Pro / Max / Team / API) | Skill + hook auto-installed in `~/.claude/` | Claude Code itself, using your logged-in session | **No** |
-| **Claude Code OAuth or Codex CLI / ChatGPT Plus** | `simplicio task` with `SIMPLICIO_MODEL=claude-cli/<m>` or `codex-cli/<m>` | Shell-out to `claude -p` / `codex exec` (subprocess uses your existing login) | **No** |
-| **API key** for any provider (Anthropic, OpenAI, OpenRouter, GLM, DeepSeek, Ollama…) | `simplicio task` standalone CLI | The provider SDK directly | **Yes** — set `SIMPLICIO_API_KEY` |
+| **Claude Code OAuth or Codex CLI / ChatGPT Plus** | `simplicio-py task` with `SIMPLICIO_MODEL=claude-cli/<m>` or `codex-cli/<m>` | Shell-out to `claude -p` / `codex exec` (subprocess uses your existing login) | **No** |
+| **API key** for any provider (Anthropic, OpenAI, OpenRouter, GLM, DeepSeek, Ollama…) | `simplicio-py task` standalone CLI | The provider SDK directly | **Yes** — set `SIMPLICIO_API_KEY` |
 
-**Most users land on Path 1.** `pip install simplicio-cli` puts the binary on PATH; the first invocation auto-installs the skill + hook in `~/.claude/` (idempotent, opt-out via `SIMPLICIO_SKIP_AUTO_INIT=1`). From that moment, every code-edit prompt you type **inside Claude Code** is silently routed through simplicio's 6-layer contract — no extra config, no key, no cost beyond your existing Claude subscription.
+**Most users land on Path 1.** `pip install simplicio-cli` puts `simplicio-py` on PATH; the first invocation auto-installs the skill + hook in `~/.claude/` (idempotent, opt-out via `SIMPLICIO_SKIP_AUTO_INIT=1`). From that moment, every code-edit prompt you type **inside Claude Code** is silently routed through simplicio's 6-layer contract — no extra config, no key, no cost beyond your existing Claude subscription.
 
-**Path 2 — subscription shell-out (zero key).** If you have a Claude Pro/Max session (`claude login`) or a ChatGPT Plus + Codex CLI session (`codex login`) and want to drive simplicio from CI, scripts, or any context **outside** Claude Code, set `SIMPLICIO_MODEL=claude-cli/<model>` or `codex-cli/<model>`. simplicio spawns the CLI as a subprocess; the call rides your existing OAuth session — no API key required. A recursion guard (`SIMPLICIO_HOOK_GUARD=1`) is injected so the inner CLI does not re-fire simplicio's own hook.
+**Path 2 — subscription shell-out (zero key).** If you have a Claude Pro/Max session (`claude login`) or a ChatGPT Plus + Codex CLI session (`codex login`) and want to drive simplicio from CI, scripts, or any context **outside** Claude Code, set `SIMPLICIO_MODEL=claude-cli/<model>` or `codex-cli/<model>`. `simplicio-py` spawns the CLI as a subprocess; the call rides your existing OAuth session — no API key required. A recursion guard (`SIMPLICIO_HOOK_GUARD=1`) is injected so the inner CLI does not re-fire the hook.
 
-**Path 3 is for environments without any logged-in CLI** — a remote server, a build runner, a notebook, a different LLM provider. You bring an API key (Anthropic, OpenRouter, OpenAI, GLM, DeepSeek, Ollama…), simplicio calls the provider directly.
+**Path 3 is for environments without any logged-in CLI** — a remote server, a build runner, a notebook, a different LLM provider. You bring an API key (Anthropic, OpenRouter, OpenAI, GLM, DeepSeek, Ollama…), `simplicio-py` calls the provider directly.
 
 #### Path 1 example — inside Claude Code
 
-After `pip install simplicio-cli && simplicio smoke` (which triggers auto-bootstrap), just type your task in Claude Code:
+After `pip install simplicio-cli && simplicio-py smoke` (which triggers auto-bootstrap), just type your task in Claude Code:
 
 ```
 hide the Delete button for non-admins in src/app/screen/screen.component.html
@@ -466,27 +466,27 @@ piggybacks on that login — no extra bill, no key to manage.
 export SIMPLICIO_MODEL=claude-cli/sonnet     # or claude-cli/opus, claude-cli/default
 unset  SIMPLICIO_API_KEY                     # explicitly: no key needed
 
-simplicio task "hide Delete button for non-admins" --stack angular \
+simplicio-py task "hide Delete button for non-admins" --stack angular \
   --target src/app/screen/screen.component.html
 
 # Option B — Codex CLI subscription (run `codex login` once)
 export SIMPLICIO_MODEL=codex-cli/gpt-5       # or codex-cli/default
-simplicio task "..." --stack angular --target ...
+simplicio-py task "..." --stack angular --target ...
 ```
 
-How it works: simplicio shells out to `claude -p "<prompt>"` (or `codex exec "<prompt>"`) as a subprocess, captures stdout, runs the test loop. The inner CLI authenticates via your existing OAuth session in `~/.claude/` or `~/.codex/`. simplicio sets `SIMPLICIO_HOOK_GUARD=1` in the subprocess env so the inner Claude Code session does **not** re-fire simplicio's own UserPromptSubmit hook (no infinite recursion).
+How it works: `simplicio-py` shells out to `claude -p "<prompt>"` (or `codex exec "<prompt>"`) as a subprocess, captures stdout, runs the test loop. The inner CLI authenticates via your existing OAuth session in `~/.claude/` or `~/.codex/`. `simplicio-py` sets `SIMPLICIO_HOOK_GUARD=1` in the subprocess env so the inner Claude Code session does **not** re-fire its own UserPromptSubmit hook (no infinite recursion).
 
-For orchestrators such as SendSprint, `simplicio task` also has a structured
+For orchestrators such as SendSprint, `simplicio-py task` also has a structured
 contract:
 
 ```bash
-simplicio task "hide Delete button for non-admins" \
+simplicio-py task "hide Delete button for non-admins" \
   --stack angular \
   --target src/app/screen/screen.component.html \
   --dry-run-task \
   --json
 
-simplicio task "front-only task" \
+simplicio-py task "front-only task" \
   --stack angular \
   --target src/app/screen/screen.component.html \
   --bound-paths "src/app/**" \
@@ -506,8 +506,8 @@ export SIMPLICIO_API_KEY=sk-or-v1-…                      # OpenRouter key
 export SIMPLICIO_MODEL=anthropic/claude-opus-4
 export SIMPLICIO_BASE_URL=https://openrouter.ai/api/v1
 
-simplicio index --stack angular                           # one-time, builds embedding cache
-simplicio task "hide Delete button for non-admins" \
+simplicio-py index --stack angular                           # one-time, builds embedding cache
+simplicio-py task "hide Delete button for non-admins" \
   --stack angular \
   --target src/app/screen/screen.component.html \
   --criteria "- no admin perm: button absent from DOM
@@ -522,20 +522,20 @@ Provider-agnostic — see [Configure](#configure--any-llm-nothing-hardcoded) for
 
 #### Path 1 deep-dive — auto-activation in Claude Code
 
-`pip install` puts `simplicio` on your PATH. To make Claude Code
+`pip install` puts `simplicio-py` on your PATH. To make Claude Code
 **automatically** route code-edit tasks through simplicio, a skill + hook
 need to land in `~/.claude/`.
 
-**Zero-step path (recommended).** The first time you run *any* `simplicio`
+**Zero-step path (recommended).** The first time you run *any* `simplicio-py`
 command after install, if Claude Code is present (`~/.claude/` exists) and
-the hook is missing, simplicio installs both for you and prints one stderr
+the hook is missing, `simplicio-py` installs both for you and prints one stderr
 line. PEP 517 wheels can't execute code on `pip install`, so this is the
 closest equivalent that works on every machine.
 
 ```bash
 pip install simplicio-cli
-simplicio smoke         # ← first call also installs skill + hook (idempotent)
-# stderr: "simplicio: auto-activation installed in Claude Code …"
+simplicio-py smoke         # ← first call also installs skill + hook (idempotent)
+# stderr: "simplicio-py: auto-activation installed in Claude Code …"
 ```
 
 Opt out before the first call:
@@ -547,9 +547,9 @@ export SIMPLICIO_SKIP_AUTO_INIT=1
 **Explicit path.** Same effect, no auto-magic:
 
 ```bash
-simplicio init                 # idempotent
-simplicio init --dry-run       # preview only
-simplicio init --claude-home <path>   # override target dir
+simplicio-py init                 # idempotent
+simplicio-py init --dry-run       # preview only
+simplicio-py init --claude-home <path>   # override target dir
 ```
 
 Either way, two files land in `~/.claude/`:
@@ -557,7 +557,7 @@ Either way, two files land in `~/.claude/`:
 | File | Purpose |
 |---|---|
 | `~/.claude/skills/simplicio-cli/SKILL.md` | Skill the agent matches by description when your prompt looks like a code edit |
-| `~/.claude/hooks/simplicio-userpromptsubmit.sh` + entry in `~/.claude/settings.json` | UserPromptSubmit hook that runs `simplicio detect` on every prompt and injects a hint when the heuristic catches a code-edit task the skill could miss |
+| `~/.claude/hooks/simplicio-userpromptsubmit.sh` + entry in `~/.claude/settings.json` | UserPromptSubmit hook that runs `simplicio-py detect` on every prompt and injects a hint when the heuristic catches a code-edit task the skill could miss |
 
 A backup of your previous `settings.json` is written to `settings.json.bak`
 before any merge.
@@ -568,13 +568,13 @@ After install, every prompt you type in Claude Code flows through two layers:
 
 1. **Skill layer (semantic).** Claude reads the SKILL.md description. When
    your prompt looks like a programming task ("add X to Y.tsx", "fix the auth
-   bug in middleware.py"), Claude considers using `simplicio task` instead of
+   bug in middleware.py"), Claude considers using `simplicio-py task` instead of
    writing code directly.
-2. **Hook layer (deterministic).** Every prompt fires `simplicio detect` via
+2. **Hook layer (deterministic).** Every prompt fires `simplicio-py detect` via
    the UserPromptSubmit hook. The classifier scores the prompt (verbs + file
    extensions + code nouns − read-only cues). Score ≥ 3 → it emits a
    `[SIMPLICIO_PROMPT_HINT]` block on stderr. Claude sees the hint alongside
-   your prompt — a hard nudge toward `simplicio task <prompt> <repo>`.
+   your prompt — a hard nudge toward `simplicio-py task <prompt> <repo>`.
 
 The layers are complementary. Skill = "Claude *might* pick simplicio". Hook
 = "Claude *sees* the hint regardless".
@@ -590,11 +590,11 @@ user prompt. UserPromptSubmit is the right pre-hook for routing decisions.
 
 | Goal | How |
 |---|---|
-| Block the auto-bootstrap | `export SIMPLICIO_SKIP_AUTO_INIT=1` before the first `simplicio` call |
+| Block the auto-bootstrap | `export SIMPLICIO_SKIP_AUTO_INIT=1` before the first `simplicio-py` call |
 | Disable hook permanently | Delete `~/.claude/hooks/simplicio-userpromptsubmit.sh` and its entry in `~/.claude/settings.json` |
-| Re-install / repair | `simplicio init` (idempotent — won't double-write) |
-| Preview without writing | `simplicio init --dry-run` |
-| Skill-only (no hook) | Copy `.skills/simplicio-cli/SKILL.md` to `~/.claude/skills/simplicio-cli/SKILL.md` manually, skip `simplicio init` |
+| Re-install / repair | `simplicio-py init` (idempotent — won't double-write) |
+| Preview without writing | `simplicio-py init --dry-run` |
+| Skill-only (no hook) | Copy `.skills/simplicio-cli/SKILL.md` to `~/.claude/skills/simplicio-cli/SKILL.md` manually, skip `simplicio-py init` |
 
 ---
 
@@ -610,7 +610,7 @@ user prompt. UserPromptSubmit is the right pre-hook for routing decisions.
 | GLM (z.ai) | `glm-4.6` | `https://api.z.ai/api/paas/v4` |
 | DeepSeek | `deepseek-chat` | `https://api.deepseek.com` |
 | OpenAI | `gpt-4.1` | `https://api.openai.com/v1` |
-| Local (llama.cpp) | `local-llama/default` | *(leave unset)* |
+| Local (llama.cpp) | `openbmb/minicpm5:latest` | *(leave unset)* |
 | Anthropic native | `claude-opus-4-7` | *(leave unset)* |
 
 If `SIMPLICIO_BASE_URL` is unset and the key is `ANTHROPIC_API_KEY`, it uses the
@@ -618,7 +618,7 @@ native Anthropic SDK. Otherwise it uses an OpenAI-compatible client pointed at
 your `base_url` — so **any** OpenAI-like provider works without code changes.
 
 ```bash
-simplicio smoke      # prints provider config + one test call
+simplicio-py smoke      # prints provider config + one test call
 ```
 
 #### Path 4 — local llama.cpp GGUF default
@@ -626,14 +626,14 @@ simplicio smoke      # prints provider config + one test call
 When **no provider is configured** (`SIMPLICIO_MODEL` and
 `SIMPLICIO_BASE_URL` both unset), simplicio runs the in-process
 [`llama-cpp-python`](https://github.com/abetlen/llama-cpp-python) backend with
-`local-llama/default`, currently
-`bartowski/Qwen_Qwen3.5-2B-GGUF::Qwen_Qwen3.5-2B-Q6_K.gguf`.
+`openbmb/minicpm5:latest`, backed by
+`openbmb/MiniCPM5-1B-GGUF::MiniCPM5-1B-Q4_K_M.gguf`.
 
 ```bash
 pip install 'simplicio-cli[local]'          # pulls llama-cpp-python + huggingface-hub
-simplicio doctor --install                  # downloads/validates the default GGUF
+simplicio-py doctor --install                  # downloads/validates the default GGUF
 
-simplicio task "add input validation to createUser" \
+simplicio-py task "add input validation to createUser" \
   --target src/users.ts --local              # forces local llama.cpp
 
 # the GGUF is fetched once from the Hugging Face Hub, then reused
@@ -642,17 +642,24 @@ simplicio task "add input validation to createUser" \
 Explicit routes (override the default model/weights):
 
 ```bash
-SIMPLICIO_MODEL=local-llama/default                                  # Qwen_Qwen3.5-2B-Q6_K.gguf default
-SIMPLICIO_MODEL=local-llama/bartowski/Qwen_Qwen3.5-2B-GGUF::Qwen_Qwen3.5-2B-Q6_K.gguf
+SIMPLICIO_MODEL=openbmb/minicpm5:latest                              # MiniCPM5-1B-Q4_K_M.gguf default
+SIMPLICIO_MODEL=local-llama/default                                  # backward-compatible alias
+SIMPLICIO_MODEL=local-llama/openbmb/MiniCPM5-1B-GGUF::MiniCPM5-1B-Q4_K_M.gguf
 SIMPLICIO_MODEL=local-llama//models/my-model.gguf                    # direct local path
 SIMPLICIO_LOCAL_MODEL_PATH=/models/my-model.gguf                     # always wins
 ```
 
 Tuning knobs (all optional): `SIMPLICIO_LOCAL_CTX` (context window, default
-`8192`), `SIMPLICIO_LOCAL_THREADS`, `SIMPLICIO_LOCAL_GPU_LAYERS` (offload to GPU,
-default `0`), `SIMPLICIO_LOCAL_MAX_TOKENS` (generation cap),
-`SIMPLICIO_LOCAL_TEMP` (default `0.1`), `SIMPLICIO_LOCAL_MODEL_REPO` /
-`SIMPLICIO_LOCAL_MODEL_FILE`.
+`2048`, clamped by `SIMPLICIO_LOCAL_CTX_MAX`, default `4096`),
+`SIMPLICIO_LOCAL_THREADS` (default and cap `4` via
+`SIMPLICIO_LOCAL_THREADS_MAX`), `SIMPLICIO_LOCAL_GPU_LAYERS` (offload to GPU,
+default `0`), `SIMPLICIO_LOCAL_BATCH` (default/cap `128`),
+`SIMPLICIO_LOCAL_UBATCH` (default/cap `32`), `SIMPLICIO_LOCAL_MAX_TOKENS`
+(generation cap, default `512`, clamped by `SIMPLICIO_LOCAL_MAX_TOKENS_CAP`,
+default `2048`), `SIMPLICIO_LOCAL_TEMP` (default `0.1`),
+`SIMPLICIO_LOCAL_MODEL_REPO` / `SIMPLICIO_LOCAL_MODEL_FILE`. The runtime keeps
+`mmap` enabled and `mlock` disabled so `llama.cpp` does not accidentally
+over-allocate RAM.
 
 #### The pipeline (both paths)
 
@@ -683,10 +690,10 @@ prompt. No key needed.
 **"I want to run it in CI / a script / outside Claude Code."** Path 2. Get an
 API key from any of the providers above (OpenRouter is the cheapest way to
 try multiple models behind one key), set `SIMPLICIO_API_KEY` +
-`SIMPLICIO_MODEL` + optional `SIMPLICIO_BASE_URL`, run `simplicio task ...`.
+`SIMPLICIO_MODEL` + optional `SIMPLICIO_BASE_URL`, run `simplicio-py task ...`.
 
 **"How do I load `.env.local` safely before running a local API?"** Use
-`eval "$(simplicio env-export .env.local)"` instead of `source .env.local`.
+`eval "$(simplicio-py env-export .env.local)"` instead of `source .env.local`.
 This preserves values with semicolons, such as PostgreSQL connection strings,
 without executing the dotenv file as shell code.
 
@@ -698,7 +705,7 @@ tracked, not shipped.
 
 **"Will Claude Code use simplicio for *every* prompt now?"** No. The skill
 only triggers on prompts that look like code edits (the description is
-specific). The hook fires `simplicio detect` on every prompt but only emits
+specific). The hook fires `simplicio-py detect` on every prompt but only emits
 a hint when the deterministic classifier scores ≥ 3 (verbs + file extensions
 + code nouns − read-only cues). "What does this function do?" gets no
 nudge. "Add a delete confirmation to UserList.tsx" does.
@@ -738,7 +745,7 @@ block, TEST block, contract-state words. Full numbers in [`bench/results.md`](be
 #### Full harness (your real project, your real tests)
 
 ```bash
-simplicio bench --cases bench/cases.json --stack angular
+simplicio-py bench --cases bench/cases.json --stack angular
 ```
 
 Runs each case two ways and runs **your real test command** (e.g. `ng test
@@ -852,7 +859,7 @@ bench/
   run_offline.py  # stdlib-only multi-model benchmark
   cases.json      # your benchmark tasks
   cases_offline.json
-  results.md      # filled by `simplicio bench` / `run_offline.py`
+  results.md      # filled by `simplicio-py bench` / `run_offline.py`
   charts/         # SVG: overall, delta, by_case, by_stack
 ```
 

{simplicio_cli-0.5.17 → simplicio_cli-0.6.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "simplicio-cli"
-version = "0.5.17"
+version = "0.6.0"
 description = "Portable task-to-code pipeline that works with any LLM. Turn a one-line task into a verified code change — diff + test + verify loop. +55 pts on a 156-check benchmark, 21% faster, ~same tokens."
 readme = "README.md"
 license = { text = "MIT" }
@@ -40,23 +40,23 @@ classifiers = [
   "Topic :: Scientific/Engineering :: Artificial Intelligence",
 ]
 dependencies = [
-  "sentence-transformers>=5.1.2",
-  "numpy>=2.0.2",
-  "anthropic>=0.105.2",
-  "openai>=2.38.0",
-  "simplicio-mapper>=0.7.3",
-  "simplicio-prompt>=1.13.3",
+  "sentence-transformers>=5.6.0",
+  "numpy>=2.5.0",
+  "anthropic>=0.112.0",
+  "openai>=2.44.0",
+  "simplicio-mapper>=0.9.0",
+  "simplicio-prompt>=1.14.1",
   "httpx>=0.28.1",
-  "orjson>=3.11.5",
+  "orjson>=3.11.9",
   "diskcache>=5.6.3",
   "libcst>=1.8.6",
 ]
 
 [project.optional-dependencies]
-bench = ["fpdf2>=2.8.4"]
+bench = ["fpdf2>=2.8.7"]
 # Offline fallback in-process inference. Pulls the llama.cpp Python bindings
-# plus huggingface-hub to fetch Qwen_Qwen3.5-2B-Q6_K.gguf.
-local = ["llama-cpp-python>=0.3.23", "huggingface-hub>=1.8.0"]
+# plus huggingface-hub to fetch MiniCPM5-1B-Q4_K_M.gguf.
+local = ["llama-cpp-python>=0.3.32", "huggingface-hub>=1.21.0"]
 
 [project.urls]
 Homepage = "https://github.com/wesleysimplicio/simplicio-cli"
@@ -65,7 +65,9 @@ Issues = "https://github.com/wesleysimplicio/simplicio-cli/issues"
 Changelog = "https://github.com/wesleysimplicio/simplicio-cli/releases"
 
 [project.scripts]
-simplicio = "simplicio.cli:main"
+simplicio-cli = "simplicio.cli:main"
+simplicio-py = "simplicio.cli:main"
+simplicio-dev-cli = "simplicio.cli:main"
 
 [build-system]
 requires = ["setuptools>=68", "wheel"]

simplicio_cli-0.6.0/simplicio/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.6.0"