prompture 0.0.46.dev1__tar.gz → 0.0.47__tar.gz

prompture-0.0.47/.claude/skills/add-driver/SKILL.md

@@ -0,0 +1,221 @@
+---
+name: add-driver
+description: Scaffold a new LLM provider driver for Prompture. Creates sync + async driver classes, registers them in the driver registry, adds settings, env template, setup.py extras, package exports, discovery integration, and models.dev pricing. Use when adding support for a new LLM provider.
+metadata:
+  author: prompture
+  version: "2.0"
+---
+
+# Add a New LLM Driver
+
+Scaffolds all files needed to integrate a new LLM provider into Prompture.
+
+## Before Starting
+
+Ask the user for:
+- **Provider name** (lowercase, used as registry key and `provider/model` prefix)
+- **SDK package name** on PyPI and minimum version (or `requests`/`httpx` for raw HTTP)
+- **Default model ID**
+- **Authentication** — API key env var name, endpoint URL, or both
+- **API compatibility** — OpenAI-compatible (`/v1/chat/completions`), custom SDK, or proprietary HTTP
+- **Lazy or eager import** — lazy if SDK is optional, eager if it's in `install_requires`
+
+Also look up the provider on [models.dev](https://models.dev) to determine:
+- **models.dev provider name** (e.g., `"anthropic"` for Claude, `"xai"` for Grok, `"moonshotai"` for Moonshot)
+- **Whether models.dev has entries** — if yes, pricing comes from models.dev live data (set `MODEL_PRICING = {}`). If no, add hardcoded pricing.
+
+## Files to Create or Modify (11 total)
+
+### 1. NEW: `prompture/drivers/{provider}_driver.py` (sync driver)
+
+See [references/driver-template.md](references/driver-template.md) for the full skeleton.
+
+Key rules:
+- Subclass `CostMixin, Driver` (NOT just `Driver`)
+- Set class-level capability flags: `supports_json_mode`, `supports_json_schema`, `supports_tool_use`, `supports_streaming`, `supports_vision`, `supports_messages`
+- Use `self._get_model_config(provider, model)` to get per-model `tokens_param` and `supports_temperature` from models.dev
+- Use `self._calculate_cost(provider, model, prompt_tokens, completion_tokens)` — do NOT manually compute costs
+- Use `self._validate_model_capabilities(provider, model, ...)` before API calls to warn about unsupported features
+- If models.dev has this provider's data, set `MODEL_PRICING = {}` (empty — pricing comes live from models.dev)
+- `generate()` returns `{"text": str, "meta": dict}`
+- `meta` MUST contain: `prompt_tokens`, `completion_tokens`, `total_tokens`, `cost`, `raw_response`, `model_name`
+- Implement `generate_messages()`, `generate_messages_with_tools()`, and `generate_messages_stream()` for full feature support
+- Optional SDK: wrap import in try/except, raise `ImportError` pointing to `pip install prompture[{provider}]`
+
+### 2. NEW: `prompture/drivers/async_{provider}_driver.py` (async driver)
+
+Mirror of the sync driver using `AsyncDriver` base class:
+- Subclass `CostMixin, AsyncDriver`
+- Same capability flags as the sync driver
+- Share `MODEL_PRICING` from the sync driver: `MODEL_PRICING = {Provider}Driver.MODEL_PRICING`
+- Use `httpx.AsyncClient` for HTTP calls (or async SDK methods)
+- All generate methods are `async def`
+- Streaming returns `AsyncIterator[dict[str, Any]]`
+
+### 3. `prompture/drivers/__init__.py`
+
+- Add sync import: `from .{provider}_driver import {Provider}Driver`
+- Add async import: `from .async_{provider}_driver import Async{Provider}Driver`
+- Register sync driver with `register_driver()`:
+  ```python
+  register_driver(
+      "{provider}",
+      lambda model=None: {Provider}Driver(
+          api_key=settings.{provider}_api_key,
+          model=model or settings.{provider}_model,
+      ),
+      overwrite=True,
+  )
+  ```
+- Add `"{Provider}Driver"` and `"Async{Provider}Driver"` to `__all__`
+
+### 4. `prompture/__init__.py`
+
+- Add `{Provider}Driver` to the `.drivers` import line
+- Add `"{Provider}Driver"` to `__all__` under `# Drivers`
+
+### 5. `prompture/settings.py`
+
+Add inside `Settings` class:
+```python
+# {Provider}
+{provider}_api_key: Optional[str] = None
+{provider}_model: str = "default-model"
+# Add endpoint if the provider supports custom endpoints:
+# {provider}_endpoint: str = "https://api.example.com/v1"
+```
+
+### 6. `prompture/discovery.py`
+
+Two changes required:
+
+**a) Add to `provider_classes` dict and configuration check:**
+- Import the driver class at the top of the file
+- Add to `provider_classes`: `"{provider}": {Provider}Driver`
+- Add configuration check in the `is_configured` block:
+  ```python
+  elif provider == "{provider}":
+      if settings.{provider}_api_key or os.getenv("{PROVIDER}_API_KEY"):
+          is_configured = True
+  ```
+  For local/endpoint-only providers (like ollama), use endpoint presence instead.
+
+**b) This ensures `get_available_models()` returns the provider's models** from both:
+- Static detection: `MODEL_PRICING` keys (or empty if pricing is from models.dev)
+- models.dev enrichment: via `PROVIDER_MAP` in `model_rates.py` (see step 7)
+
+### 7. `prompture/model_rates.py` — `PROVIDER_MAP`
+
+If models.dev has this provider's data, add the mapping:
+```python
+PROVIDER_MAP: dict[str, str] = {
+    ...
+    "{provider}": "{models_dev_name}",  # e.g., "moonshot": "moonshotai"
+}
+```
+
+This enables:
+- **Live pricing** via `get_model_rates()` — used by `CostMixin._calculate_cost()`
+- **Capability metadata** via `get_model_capabilities()` — used by `_get_model_config()` and `_validate_model_capabilities()`
+- **Model discovery** via `get_all_provider_models()` — called by `discovery.py` to list all available models
+
+To find the correct models.dev name, check: `https://models.dev/{models_dev_name}`
+
+If models.dev does NOT have this provider, skip this step. The driver will use hardcoded `MODEL_PRICING` for costs and return `None` for capabilities.
+
+### 8. `setup.py` / `pyproject.toml`
+
+If optional: add `"{provider}": ["{sdk}>={version}"]` to `extras_require`.
+If required: add to `install_requires`.
+
+### 9. `.env.copy`
+
+Add section:
+```
+# {Provider} Configuration
+{PROVIDER}_API_KEY=your-api-key-here
+{PROVIDER}_MODEL=default-model
+```
+
+### 10. `CLAUDE.md`
+
+Add `{provider}` to the driver list in the Module Layout bullet.
+
+### 11. OPTIONAL: `examples/{provider}_example.py`
+
+Follow the existing example pattern (see `grok_example.py` or `groq_example.py`):
+- Two extraction examples: default instruction + custom instruction
+- Show different models if available
+- Print JSON output and token usage statistics
+
+## Important: Reasoning Model Handling
+
+If the provider has reasoning models (models with `reasoning: true` on models.dev):
+- Check `caps.is_reasoning` before sending `response_format` — reasoning models often don't support it
+- Handle `reasoning_content` field in responses (both regular and streaming)
+- Some reasoning models don't support `temperature` — respect `supports_temperature` from `_get_model_config()`
+
+Example pattern (see `moonshot_driver.py`):
+```python
+if options.get("json_mode"):
+    from ..model_rates import get_model_capabilities
+
+    caps = get_model_capabilities("{provider}", model)
+    is_reasoning = caps is not None and caps.is_reasoning is True
+    model_supports_structured = (
+        caps is None or caps.supports_structured_output is not False
+    ) and not is_reasoning
+
+    if model_supports_structured:
+        # Send response_format
+        ...
+```
+
+## How models.dev Integration Works
+
+```
+User calls extract_and_jsonify("moonshot/kimi-k2.5", ...)
+    │
+    ├─► core.py checks driver.supports_json_mode → decides json_mode
+    │
+    ├─► driver._get_model_config("moonshot", "kimi-k2.5")
+    │       └─► model_rates.get_model_capabilities("moonshot", "kimi-k2.5")
+    │               └─► PROVIDER_MAP["moonshot"] → "moonshotai"
+    │               └─► models.dev data["moonshotai"]["models"]["kimi-k2.5"]
+    │               └─► Returns: supports_temperature, is_reasoning, context_window, etc.
+    │
+    ├─► driver._calculate_cost("moonshot", "kimi-k2.5", tokens...)
+    │       └─► model_rates.get_model_rates("moonshot", "kimi-k2.5")
+    │               └─► Same lookup → returns {input: 0.6, output: 3.0} per 1M tokens
+    │
+    └─► discovery.get_available_models()
+            └─► Iterates PROVIDER_MAP → get_all_provider_models("moonshotai")
+            └─► Returns all model IDs under the provider
+```
+
+## Model Name Resolution
+
+Model names are **always provider-scoped**. The format is `"provider/model_id"`.
+
+- `get_driver_for_model("openrouter/qwen-2.5")` → looks up `"openrouter"` in the driver registry
+- `get_model_capabilities("openrouter", "qwen-2.5")` → looks in models.dev under `data["openrouter"]["models"]["qwen-2.5"]`
+- `get_model_capabilities("modelscope", "qwen-2.5")` → looks in models.dev under `data["modelscope"]["models"]["qwen-2.5"]`
+
+The same model ID under different providers is **not ambiguous** — each provider has its own namespace in both the driver registry and models.dev data.
+
+## Verification
+
+```bash
+# Import check
+python -c "from prompture import {Provider}Driver; print('OK')"
+python -c "from prompture.drivers import Async{Provider}Driver; print('OK')"
+
+# Registry check
+python -c "from prompture.drivers import get_driver_for_model; d = get_driver_for_model('{provider}/test'); print(type(d).__name__, d.model)"
+
+# Discovery check
+python -c "from prompture import get_available_models; ms = [m for m in get_available_models() if m.startswith('{provider}/')]; print(f'Found {{len(ms)}} models'); print(ms[:5])"
+
+# Run tests
+pytest tests/ -x -q
+```

prompture-0.0.47/.claude/skills/add-driver/references/driver-template.md

@@ -0,0 +1,364 @@
+# Driver Template
+
+Every Prompture driver follows this skeleton. The sync driver uses `requests`,
+the async driver uses `httpx`.
+
+## Sync Driver — `prompture/drivers/{provider}_driver.py`
+
+```python
+"""{Provider} driver implementation.
+Requires the `requests` package. Uses {PROVIDER}_API_KEY env var.
+
+All pricing comes from models.dev (provider: "{models_dev_name}") — no hardcoded pricing.
+"""
+
+import json
+import logging
+import os
+from collections.abc import Iterator
+from typing import Any
+
+import requests
+
+from ..cost_mixin import CostMixin, prepare_strict_schema
+from ..driver import Driver
+
+logger = logging.getLogger(__name__)
+
+
+class {Provider}Driver(CostMixin, Driver):
+    supports_json_mode = True
+    supports_json_schema = True
+    supports_tool_use = True
+    supports_streaming = True
+    supports_vision = False  # set True if the provider supports image input
+    supports_messages = True
+
+    # All pricing resolved live from models.dev (provider: "{models_dev_name}")
+    # If models.dev does NOT have this provider, add hardcoded pricing:
+    #   MODEL_PRICING = {
+    #       "model-name": {"prompt": 0.001, "completion": 0.002},
+    #   }
+    MODEL_PRICING: dict[str, dict[str, Any]] = {}
+
+    def __init__(
+        self,
+        api_key: str | None = None,
+        model: str = "default-model",
+        endpoint: str = "https://api.example.com/v1",
+    ):
+        self.api_key = api_key or os.getenv("{PROVIDER}_API_KEY")
+        if not self.api_key:
+            raise ValueError("{Provider} API key not found. Set {PROVIDER}_API_KEY env var.")
+
+        self.model = model
+        self.base_url = endpoint.rstrip("/")
+        self.headers = {
+            "Authorization": f"Bearer {self.api_key}",
+            "Content-Type": "application/json",
+        }
+
+    def generate(self, prompt: str, options: dict[str, Any]) -> dict[str, Any]:
+        messages = [{"role": "user", "content": prompt}]
+        return self._do_generate(messages, options)
+
+    def generate_messages(self, messages: list[dict[str, str]], options: dict[str, Any]) -> dict[str, Any]:
+        return self._do_generate(messages, options)
+
+    def _do_generate(self, messages: list[dict[str, str]], options: dict[str, Any]) -> dict[str, Any]:
+        model = options.get("model", self.model)
+
+        # Per-model config from models.dev (tokens_param, supports_temperature, etc.)
+        model_config = self._get_model_config("{provider}", model)
+        tokens_param = model_config["tokens_param"]
+        supports_temperature = model_config["supports_temperature"]
+
+        # Validate capabilities (logs warnings if model doesn't support requested features)
+        self._validate_model_capabilities(
+            "{provider}",
+            model,
+            using_json_schema=bool(options.get("json_schema")),
+        )
+
+        opts = {"temperature": 1.0, "max_tokens": 512, **options}
+
+        data: dict[str, Any] = {
+            "model": model,
+            "messages": messages,
+        }
+        data[tokens_param] = opts.get("max_tokens", 512)
+
+        if supports_temperature and "temperature" in opts:
+            data["temperature"] = opts["temperature"]
+
+        # Native JSON mode — check per-model capabilities before sending response_format
+        if options.get("json_mode"):
+            from ..model_rates import get_model_capabilities
+
+            caps = get_model_capabilities("{provider}", model)
+            is_reasoning = caps is not None and caps.is_reasoning is True
+            model_supports_structured = (
+                caps is None or caps.supports_structured_output is not False
+            ) and not is_reasoning
+
+            if model_supports_structured:
+                json_schema = options.get("json_schema")
+                if json_schema:
+                    schema_copy = prepare_strict_schema(json_schema)
+                    data["response_format"] = {
+                        "type": "json_schema",
+                        "json_schema": {
+                            "name": "extraction",
+                            "strict": True,
+                            "schema": schema_copy,
+                        },
+                    }
+                else:
+                    data["response_format"] = {"type": "json_object"}
+
+        try:
+            response = requests.post(
+                f"{self.base_url}/chat/completions",
+                headers=self.headers,
+                json=data,
+                timeout=120,
+            )
+            response.raise_for_status()
+            resp = response.json()
+        except requests.exceptions.HTTPError as e:
+            raise RuntimeError(f"{Provider} API request failed: {e!s}") from e
+        except requests.exceptions.RequestException as e:
+            raise RuntimeError(f"{Provider} API request failed: {e!s}") from e
+
+        usage = resp.get("usage", {})
+        prompt_tokens = usage.get("prompt_tokens", 0)
+        completion_tokens = usage.get("completion_tokens", 0)
+        total_tokens = usage.get("total_tokens", 0)
+
+        # Cost calculated from models.dev live rates, falling back to MODEL_PRICING
+        total_cost = self._calculate_cost("{provider}", model, prompt_tokens, completion_tokens)
+
+        meta = {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": total_tokens,
+            "cost": round(total_cost, 6),
+            "raw_response": resp,
+            "model_name": model,
+        }
+
+        message = resp["choices"][0]["message"]
+        text = message.get("content") or ""
+
+        # Reasoning models may return content in reasoning_content when content is empty
+        if not text and message.get("reasoning_content"):
+            text = message["reasoning_content"]
+
+        return {"text": text, "meta": meta}
+
+    # ------------------------------------------------------------------
+    # Tool use
+    # ------------------------------------------------------------------
+
+    def generate_messages_with_tools(
+        self,
+        messages: list[dict[str, Any]],
+        tools: list[dict[str, Any]],
+        options: dict[str, Any],
+    ) -> dict[str, Any]:
+        """Generate a response that may include tool calls."""
+        model = options.get("model", self.model)
+        model_config = self._get_model_config("{provider}", model)
+        tokens_param = model_config["tokens_param"]
+        supports_temperature = model_config["supports_temperature"]
+
+        self._validate_model_capabilities("{provider}", model, using_tool_use=True)
+
+        opts = {"temperature": 1.0, "max_tokens": 512, **options}
+
+        data: dict[str, Any] = {
+            "model": model,
+            "messages": messages,
+            "tools": tools,
+        }
+        data[tokens_param] = opts.get("max_tokens", 512)
+
+        if supports_temperature and "temperature" in opts:
+            data["temperature"] = opts["temperature"]
+
+        if "tool_choice" in options:
+            data["tool_choice"] = options["tool_choice"]
+
+        try:
+            response = requests.post(
+                f"{self.base_url}/chat/completions",
+                headers=self.headers,
+                json=data,
+                timeout=120,
+            )
+            response.raise_for_status()
+            resp = response.json()
+        except requests.exceptions.HTTPError as e:
+            raise RuntimeError(f"{Provider} API request failed: {e!s}") from e
+        except requests.exceptions.RequestException as e:
+            raise RuntimeError(f"{Provider} API request failed: {e!s}") from e
+
+        usage = resp.get("usage", {})
+        prompt_tokens = usage.get("prompt_tokens", 0)
+        completion_tokens = usage.get("completion_tokens", 0)
+        total_tokens = usage.get("total_tokens", 0)
+        total_cost = self._calculate_cost("{provider}", model, prompt_tokens, completion_tokens)
+
+        meta = {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": total_tokens,
+            "cost": round(total_cost, 6),
+            "raw_response": resp,
+            "model_name": model,
+        }
+
+        choice = resp["choices"][0]
+        text = choice["message"].get("content") or ""
+        stop_reason = choice.get("finish_reason")
+
+        tool_calls_out: list[dict[str, Any]] = []
+        for tc in choice["message"].get("tool_calls", []):
+            try:
+                args = json.loads(tc["function"]["arguments"])
+            except (json.JSONDecodeError, TypeError):
+                args = {}
+            tool_calls_out.append({
+                "id": tc["id"],
+                "name": tc["function"]["name"],
+                "arguments": args,
+            })
+
+        return {
+            "text": text,
+            "meta": meta,
+            "tool_calls": tool_calls_out,
+            "stop_reason": stop_reason,
+        }
+
+    # ------------------------------------------------------------------
+    # Streaming
+    # ------------------------------------------------------------------
+
+    def generate_messages_stream(
+        self,
+        messages: list[dict[str, Any]],
+        options: dict[str, Any],
+    ) -> Iterator[dict[str, Any]]:
+        """Yield response chunks via streaming API."""
+        model = options.get("model", self.model)
+        model_config = self._get_model_config("{provider}", model)
+        tokens_param = model_config["tokens_param"]
+        supports_temperature = model_config["supports_temperature"]
+
+        opts = {"temperature": 1.0, "max_tokens": 512, **options}
+
+        data: dict[str, Any] = {
+            "model": model,
+            "messages": messages,
+            "stream": True,
+            "stream_options": {"include_usage": True},
+        }
+        data[tokens_param] = opts.get("max_tokens", 512)
+
+        if supports_temperature and "temperature" in opts:
+            data["temperature"] = opts["temperature"]
+
+        response = requests.post(
+            f"{self.base_url}/chat/completions",
+            headers=self.headers,
+            json=data,
+            stream=True,
+            timeout=120,
+        )
+        response.raise_for_status()
+
+        full_text = ""
+        prompt_tokens = 0
+        completion_tokens = 0
+
+        for line in response.iter_lines(decode_unicode=True):
+            if not line or not line.startswith("data: "):
+                continue
+            payload = line[len("data: "):]
+            if payload.strip() == "[DONE]":
+                break
+            try:
+                chunk = json.loads(payload)
+            except json.JSONDecodeError:
+                continue
+
+            usage = chunk.get("usage")
+            if usage:
+                prompt_tokens = usage.get("prompt_tokens", 0)
+                completion_tokens = usage.get("completion_tokens", 0)
+
+            choices = chunk.get("choices", [])
+            if choices:
+                delta = choices[0].get("delta", {})
+                content = delta.get("content") or ""
+                # Reasoning models stream thinking via reasoning_content
+                if not content:
+                    content = delta.get("reasoning_content") or ""
+                if content:
+                    full_text += content
+                    yield {"type": "delta", "text": content}
+
+        total_tokens = prompt_tokens + completion_tokens
+        total_cost = self._calculate_cost("{provider}", model, prompt_tokens, completion_tokens)
+
+        yield {
+            "type": "done",
+            "text": full_text,
+            "meta": {
+                "prompt_tokens": prompt_tokens,
+                "completion_tokens": completion_tokens,
+                "total_tokens": total_tokens,
+                "cost": round(total_cost, 6),
+                "raw_response": {},
+                "model_name": model,
+            },
+        }
+```
+
+## Lazy Import Pattern (for optional SDKs)
+
+```python
+def __init__(self, ...):
+    self._client = None
+    # defer import
+
+def _ensure_client(self):
+    if self._client is not None:
+        return
+    try:
+        from some_sdk import Client
+    except ImportError:
+        raise ImportError(
+            "The 'some-sdk' package is required. "
+            "Install with: pip install prompture[provider]"
+        )
+    self._client = Client(api_key=self.api_key)
+```
+
+## Existing Drivers for Reference
+
+| Driver | File | SDK | Auth | models.dev |
+|--------|------|-----|------|------------|
+| OpenAI | `openai_driver.py` | `openai` | API key | `openai` |
+| Claude | `claude_driver.py` | `anthropic` | API key | `anthropic` |
+| Google | `google_driver.py` | `google-generativeai` | API key | `google` |
+| Groq | `groq_driver.py` | `groq` | API key | `groq` |
+| Grok | `grok_driver.py` | `requests` | API key | `xai` |
+| Moonshot | `moonshot_driver.py` | `requests` | API key + endpoint | `moonshotai` |
+| Z.ai | `zai_driver.py` | `requests` | API key + endpoint | `zai` |
+| ModelScope | `modelscope_driver.py` | `requests` | API key + endpoint | — |
+| OpenRouter | `openrouter_driver.py` | `requests` | API key | `openrouter` |
+| Ollama | `ollama_driver.py` | `requests` | Endpoint URL | — |
+| LM Studio | `lmstudio_driver.py` | `requests` | Endpoint URL | — |
+| AirLLM | `airllm_driver.py` | `airllm` (lazy) | None (local) | — |