prompture 0.0.33.dev2__py3-none-any.whl → 0.0.34.dev1__py3-none-any.whl

prompture/drivers/async_local_http_driver.py

@@ -0,0 +1,44 @@
+"""Async LocalHTTP driver using httpx."""
+
+from __future__ import annotations
+
+import os
+from typing import Any
+
+import httpx
+
+from ..async_driver import AsyncDriver
+
+
+class AsyncLocalHTTPDriver(AsyncDriver):
+    MODEL_PRICING = {"default": {"prompt": 0.0, "completion": 0.0}}
+
+    def __init__(self, endpoint: str | None = None, model: str = "local-model"):
+        self.endpoint = endpoint or os.getenv("LOCAL_HTTP_ENDPOINT", "http://localhost:8000/generate")
+        self.model = model
+
+    async def generate(self, prompt: str, options: dict[str, Any]) -> dict[str, Any]:
+        payload = {"prompt": prompt, "options": options}
+
+        async with httpx.AsyncClient() as client:
+            try:
+                r = await client.post(self.endpoint, json=payload, timeout=options.get("timeout", 30))
+                r.raise_for_status()
+                response_data = r.json()
+            except Exception as e:
+                raise RuntimeError(f"AsyncLocalHTTPDriver request failed: {e}") from e
+
+        if "text" in response_data and "meta" in response_data:
+            return response_data
+
+        meta = {
+            "prompt_tokens": response_data.get("prompt_tokens", 0),
+            "completion_tokens": response_data.get("completion_tokens", 0),
+            "total_tokens": response_data.get("total_tokens", 0),
+            "cost": 0.0,
+            "raw_response": response_data,
+            "model_name": options.get("model", self.model),
+        }
+
+        text = response_data.get("text") or response_data.get("response") or str(response_data)
+        return {"text": text, "meta": meta}

prompture/drivers/async_ollama_driver.py

@@ -0,0 +1,125 @@
+"""Async Ollama driver using httpx."""
+
+from __future__ import annotations
+
+import logging
+import os
+from typing import Any
+
+import httpx
+
+from ..async_driver import AsyncDriver
+
+logger = logging.getLogger(__name__)
+
+
+class AsyncOllamaDriver(AsyncDriver):
+    supports_json_mode = True
+
+    MODEL_PRICING = {"default": {"prompt": 0.0, "completion": 0.0}}
+
+    def __init__(self, endpoint: str | None = None, model: str = "llama3"):
+        self.endpoint = endpoint or os.getenv("OLLAMA_ENDPOINT", "http://localhost:11434/api/generate")
+        self.model = model
+        self.options: dict[str, Any] = {}
+
+    supports_messages = True
+
+    async def generate(self, prompt: str, options: dict[str, Any] | None = None) -> dict[str, Any]:
+        merged_options = self.options.copy()
+        if options:
+            merged_options.update(options)
+
+        payload = {
+            "prompt": prompt,
+            "model": merged_options.get("model", self.model),
+            "stream": False,
+        }
+
+        # Native JSON mode support
+        if merged_options.get("json_mode"):
+            payload["format"] = "json"
+
+        if "temperature" in merged_options:
+            payload["temperature"] = merged_options["temperature"]
+        if "top_p" in merged_options:
+            payload["top_p"] = merged_options["top_p"]
+        if "top_k" in merged_options:
+            payload["top_k"] = merged_options["top_k"]
+
+        async with httpx.AsyncClient() as client:
+            try:
+                r = await client.post(self.endpoint, json=payload, timeout=120)
+                r.raise_for_status()
+                response_data = r.json()
+            except httpx.HTTPStatusError as e:
+                raise RuntimeError(f"Ollama request failed: {e}") from e
+            except Exception as e:
+                raise RuntimeError(f"Ollama request failed: {e}") from e
+
+        prompt_tokens = response_data.get("prompt_eval_count", 0)
+        completion_tokens = response_data.get("eval_count", 0)
+        total_tokens = prompt_tokens + completion_tokens
+
+        meta = {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": total_tokens,
+            "cost": 0.0,
+            "raw_response": response_data,
+            "model_name": merged_options.get("model", self.model),
+        }
+
+        return {"text": response_data.get("response", ""), "meta": meta}
+
+    async def generate_messages(self, messages: list[dict[str, str]], options: dict[str, Any]) -> dict[str, Any]:
+        """Use Ollama's /api/chat endpoint for multi-turn conversations."""
+        merged_options = self.options.copy()
+        if options:
+            merged_options.update(options)
+
+        # Derive the chat endpoint from the generate endpoint
+        chat_endpoint = self.endpoint.replace("/api/generate", "/api/chat")
+
+        payload: dict[str, Any] = {
+            "model": merged_options.get("model", self.model),
+            "messages": messages,
+            "stream": False,
+        }
+
+        if merged_options.get("json_mode"):
+            payload["format"] = "json"
+
+        if "temperature" in merged_options:
+            payload["temperature"] = merged_options["temperature"]
+        if "top_p" in merged_options:
+            payload["top_p"] = merged_options["top_p"]
+        if "top_k" in merged_options:
+            payload["top_k"] = merged_options["top_k"]
+
+        async with httpx.AsyncClient() as client:
+            try:
+                r = await client.post(chat_endpoint, json=payload, timeout=120)
+                r.raise_for_status()
+                response_data = r.json()
+            except httpx.HTTPStatusError as e:
+                raise RuntimeError(f"Ollama chat request failed: {e}") from e
+            except Exception as e:
+                raise RuntimeError(f"Ollama chat request failed: {e}") from e
+
+        prompt_tokens = response_data.get("prompt_eval_count", 0)
+        completion_tokens = response_data.get("eval_count", 0)
+        total_tokens = prompt_tokens + completion_tokens
+
+        meta = {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": total_tokens,
+            "cost": 0.0,
+            "raw_response": response_data,
+            "model_name": merged_options.get("model", self.model),
+        }
+
+        message = response_data.get("message", {})
+        text = message.get("content", "")
+        return {"text": text, "meta": meta}

prompture/drivers/async_openai_driver.py

@@ -0,0 +1,96 @@
+"""Async OpenAI driver. Requires the ``openai`` package (>=1.0.0)."""
+
+from __future__ import annotations
+
+import os
+from typing import Any
+
+try:
+    from openai import AsyncOpenAI
+except Exception:
+    AsyncOpenAI = None
+
+from ..async_driver import AsyncDriver
+from ..cost_mixin import CostMixin
+from .openai_driver import OpenAIDriver
+
+
+class AsyncOpenAIDriver(CostMixin, AsyncDriver):
+    supports_json_mode = True
+    supports_json_schema = True
+
+    MODEL_PRICING = OpenAIDriver.MODEL_PRICING
+
+    def __init__(self, api_key: str | None = None, model: str = "gpt-4o-mini"):
+        self.api_key = api_key or os.getenv("OPENAI_API_KEY")
+        self.model = model
+        if AsyncOpenAI:
+            self.client = AsyncOpenAI(api_key=self.api_key)
+        else:
+            self.client = None
+
+    supports_messages = True
+
+    async def generate(self, prompt: str, options: dict[str, Any]) -> dict[str, Any]:
+        messages = [{"role": "user", "content": prompt}]
+        return await self._do_generate(messages, options)
+
+    async def generate_messages(self, messages: list[dict[str, str]], options: dict[str, Any]) -> dict[str, Any]:
+        return await self._do_generate(messages, options)
+
+    async def _do_generate(self, messages: list[dict[str, str]], options: dict[str, Any]) -> dict[str, Any]:
+        if self.client is None:
+            raise RuntimeError("openai package (>=1.0.0) is not installed")
+
+        model = options.get("model", self.model)
+
+        model_info = self.MODEL_PRICING.get(model, {})
+        tokens_param = model_info.get("tokens_param", "max_tokens")
+        supports_temperature = model_info.get("supports_temperature", True)
+
+        opts = {"temperature": 1.0, "max_tokens": 512, **options}
+
+        kwargs = {
+            "model": model,
+            "messages": messages,
+        }
+        kwargs[tokens_param] = opts.get("max_tokens", 512)
+
+        if supports_temperature and "temperature" in opts:
+            kwargs["temperature"] = opts["temperature"]
+
+        # Native JSON mode support
+        if options.get("json_mode"):
+            json_schema = options.get("json_schema")
+            if json_schema:
+                kwargs["response_format"] = {
+                    "type": "json_schema",
+                    "json_schema": {
+                        "name": "extraction",
+                        "strict": True,
+                        "schema": json_schema,
+                    },
+                }
+            else:
+                kwargs["response_format"] = {"type": "json_object"}
+
+        resp = await self.client.chat.completions.create(**kwargs)
+
+        usage = getattr(resp, "usage", None)
+        prompt_tokens = getattr(usage, "prompt_tokens", 0)
+        completion_tokens = getattr(usage, "completion_tokens", 0)
+        total_tokens = getattr(usage, "total_tokens", 0)
+
+        total_cost = self._calculate_cost("openai", model, prompt_tokens, completion_tokens)
+
+        meta = {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": total_tokens,
+            "cost": total_cost,
+            "raw_response": resp.model_dump(),
+            "model_name": model,
+        }
+
+        text = resp.choices[0].message.content
+        return {"text": text, "meta": meta}

prompture/drivers/async_openrouter_driver.py

@@ -0,0 +1,96 @@
+"""Async OpenRouter driver using httpx."""
+
+from __future__ import annotations
+
+import os
+from typing import Any
+
+import httpx
+
+from ..async_driver import AsyncDriver
+from ..cost_mixin import CostMixin
+from .openrouter_driver import OpenRouterDriver
+
+
+class AsyncOpenRouterDriver(CostMixin, AsyncDriver):
+    supports_json_mode = True
+
+    MODEL_PRICING = OpenRouterDriver.MODEL_PRICING
+
+    def __init__(self, api_key: str | None = None, model: str = "openai/gpt-3.5-turbo"):
+        self.api_key = api_key or os.getenv("OPENROUTER_API_KEY")
+        if not self.api_key:
+            raise ValueError("OpenRouter API key not found. Set OPENROUTER_API_KEY env var.")
+        self.model = model
+        self.base_url = "https://openrouter.ai/api/v1"
+        self.headers = {
+            "Authorization": f"Bearer {self.api_key}",
+            "HTTP-Referer": "https://github.com/jhd3197/prompture",
+            "Content-Type": "application/json",
+        }
+
+    supports_messages = True
+
+    async def generate(self, prompt: str, options: dict[str, Any]) -> dict[str, Any]:
+        messages = [{"role": "user", "content": prompt}]
+        return await self._do_generate(messages, options)
+
+    async def generate_messages(self, messages: list[dict[str, str]], options: dict[str, Any]) -> dict[str, Any]:
+        return await self._do_generate(messages, options)
+
+    async def _do_generate(self, messages: list[dict[str, str]], options: dict[str, Any]) -> dict[str, Any]:
+        model = options.get("model", self.model)
+
+        model_info = self.MODEL_PRICING.get(model, {})
+        tokens_param = model_info.get("tokens_param", "max_tokens")
+        supports_temperature = model_info.get("supports_temperature", True)
+
+        opts = {"temperature": 1.0, "max_tokens": 512, **options}
+
+        data = {
+            "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 support
+        if options.get("json_mode"):
+            data["response_format"] = {"type": "json_object"}
+
+        async with httpx.AsyncClient() as client:
+            try:
+                response = await client.post(
+                    f"{self.base_url}/chat/completions",
+                    headers=self.headers,
+                    json=data,
+                    timeout=120,
+                )
+                response.raise_for_status()
+                resp = response.json()
+            except httpx.HTTPStatusError as e:
+                error_msg = f"OpenRouter API request failed: {e!s}"
+                raise RuntimeError(error_msg) from e
+            except Exception as e:
+                raise RuntimeError(f"OpenRouter 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("openrouter", model, prompt_tokens, completion_tokens)
+
+        meta = {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": total_tokens,
+            "cost": total_cost,
+            "raw_response": resp,
+            "model_name": model,
+        }
+
+        text = resp["choices"][0]["message"]["content"]
+        return {"text": text, "meta": meta}

prompture/drivers/async_registry.py

@@ -0,0 +1,80 @@
+"""Async driver registry — mirrors the sync DRIVER_REGISTRY."""
+
+from __future__ import annotations
+
+from ..settings import settings
+from .async_airllm_driver import AsyncAirLLMDriver
+from .async_azure_driver import AsyncAzureDriver
+from .async_claude_driver import AsyncClaudeDriver
+from .async_google_driver import AsyncGoogleDriver
+from .async_grok_driver import AsyncGrokDriver
+from .async_groq_driver import AsyncGroqDriver
+from .async_lmstudio_driver import AsyncLMStudioDriver
+from .async_local_http_driver import AsyncLocalHTTPDriver
+from .async_ollama_driver import AsyncOllamaDriver
+from .async_openai_driver import AsyncOpenAIDriver
+from .async_openrouter_driver import AsyncOpenRouterDriver
+
+ASYNC_DRIVER_REGISTRY = {
+    "openai": lambda model=None: AsyncOpenAIDriver(
+        api_key=settings.openai_api_key, model=model or settings.openai_model
+    ),
+    "ollama": lambda model=None: AsyncOllamaDriver(
+        endpoint=settings.ollama_endpoint, model=model or settings.ollama_model
+    ),
+    "claude": lambda model=None: AsyncClaudeDriver(
+        api_key=settings.claude_api_key, model=model or settings.claude_model
+    ),
+    "lmstudio": lambda model=None: AsyncLMStudioDriver(
+        endpoint=settings.lmstudio_endpoint, model=model or settings.lmstudio_model
+    ),
+    "azure": lambda model=None: AsyncAzureDriver(
+        api_key=settings.azure_api_key, endpoint=settings.azure_api_endpoint, deployment_id=settings.azure_deployment_id
+    ),
+    "local_http": lambda model=None: AsyncLocalHTTPDriver(endpoint=settings.local_http_endpoint, model=model),
+    "google": lambda model=None: AsyncGoogleDriver(
+        api_key=settings.google_api_key, model=model or settings.google_model
+    ),
+    "groq": lambda model=None: AsyncGroqDriver(api_key=settings.groq_api_key, model=model or settings.groq_model),
+    "openrouter": lambda model=None: AsyncOpenRouterDriver(
+        api_key=settings.openrouter_api_key, model=model or settings.openrouter_model
+    ),
+    "grok": lambda model=None: AsyncGrokDriver(api_key=settings.grok_api_key, model=model or settings.grok_model),
+    "airllm": lambda model=None: AsyncAirLLMDriver(
+        model=model or settings.airllm_model,
+        compression=settings.airllm_compression,
+    ),
+}
+
+
+def get_async_driver(provider_name: str | None = None):
+    """Factory to get an async driver instance based on the provider name.
+
+    Uses default model from settings if not overridden.
+    """
+    provider = (provider_name or settings.ai_provider or "ollama").strip().lower()
+    if provider not in ASYNC_DRIVER_REGISTRY:
+        raise ValueError(f"Unknown provider: {provider_name}")
+    return ASYNC_DRIVER_REGISTRY[provider]()
+
+
+def get_async_driver_for_model(model_str: str):
+    """Factory to get an async driver instance based on a full model string.
+
+    Format: ``provider/model_id``
+    Example: ``"openai/gpt-4-turbo-preview"``
+    """
+    if not isinstance(model_str, str):
+        raise ValueError("Model string must be a string, got {type(model_str)}")
+
+    if not model_str:
+        raise ValueError("Model string cannot be empty")
+
+    parts = model_str.split("/", 1)
+    provider = parts[0].lower()
+    model_id = parts[1] if len(parts) > 1 else None
+
+    if provider not in ASYNC_DRIVER_REGISTRY:
+        raise ValueError(f"Unsupported provider '{provider}'")
+
+    return ASYNC_DRIVER_REGISTRY[provider](model_id)

prompture/drivers/azure_driver.py

@@ -1,17 +1,23 @@
 """Driver for Azure OpenAI Service (migrated to openai>=1.0.0).
 Requires the `openai` package.
 """
+
 import os
-from typing import Any, Dict
+from typing import Any
+
 try:
     from openai import AzureOpenAI
 except Exception:
     AzureOpenAI = None
 
+from ..cost_mixin import CostMixin
 from ..driver import Driver
 
 
-class AzureDriver(Driver):
+class AzureDriver(CostMixin, Driver):
+    supports_json_mode = True
+    supports_json_schema = True
+
     # Pricing per 1K tokens (adjust if your Azure pricing differs from OpenAI defaults)
     MODEL_PRICING = {
         "gpt-5-mini": {
@@ -82,7 +88,16 @@ class AzureDriver(Driver):
         else:
             self.client = None
 
-    def generate(self, prompt: str, options: Dict[str, Any]) -> Dict[str, Any]:
+    supports_messages = True
+
+    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]:
         if self.client is None:
             raise RuntimeError("openai package (>=1.0.0) with AzureOpenAI not installed")
 
@@ -96,13 +111,28 @@ class AzureDriver(Driver):
         # Build request kwargs
         kwargs = {
             "model": self.deployment_id,  # for Azure, use deployment name
-            "messages": [{"role": "user", "content": prompt}],
+            "messages": messages,
         }
         kwargs[tokens_param] = opts.get("max_tokens", 512)
 
         if supports_temperature and "temperature" in opts:
             kwargs["temperature"] = opts["temperature"]
 
+        # Native JSON mode support
+        if options.get("json_mode"):
+            json_schema = options.get("json_schema")
+            if json_schema:
+                kwargs["response_format"] = {
+                    "type": "json_schema",
+                    "json_schema": {
+                        "name": "extraction",
+                        "strict": True,
+                        "schema": json_schema,
+                    },
+                }
+            else:
+                kwargs["response_format"] = {"type": "json_object"}
+
         resp = self.client.chat.completions.create(**kwargs)
 
         # Extract usage
@@ -111,17 +141,8 @@ class AzureDriver(Driver):
         completion_tokens = getattr(usage, "completion_tokens", 0)
         total_tokens = getattr(usage, "total_tokens", 0)
 
-        # Calculate cost — try live rates first (per 1M tokens), fall back to hardcoded (per 1K tokens)
-        from ..model_rates import get_model_rates
-        live_rates = get_model_rates("azure", model)
-        if live_rates:
-            prompt_cost = (prompt_tokens / 1_000_000) * live_rates["input"]
-            completion_cost = (completion_tokens / 1_000_000) * live_rates["output"]
-        else:
-            model_pricing = self.MODEL_PRICING.get(model, {"prompt": 0, "completion": 0})
-            prompt_cost = (prompt_tokens / 1000) * model_pricing["prompt"]
-            completion_cost = (completion_tokens / 1000) * model_pricing["completion"]
-        total_cost = prompt_cost + completion_cost
+        # Calculate cost via shared mixin
+        total_cost = self._calculate_cost("azure", model, prompt_tokens, completion_tokens)
 
         # Standardized meta object
         meta = {