File size: 11,999 Bytes
40ee6b4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
"""
LM Studio local LLM client adapter.

Implements the LLMClient protocol for LM Studio's OpenAI-compatible API.
Designed for running local models with configurable endpoint.
"""

import json
import logging
from collections.abc import AsyncIterator
from typing import Any

import httpx

from .base import BaseLLMClient, LLMResponse
from .exceptions import (
    LLMClientError,
    LLMConnectionError,
    LLMResponseParseError,
    LLMServerError,
    LLMStreamError,
    LLMTimeoutError,
)

logger = logging.getLogger(__name__)


class LMStudioClient(BaseLLMClient):
    """
    LM Studio local server client.

    LM Studio provides an OpenAI-compatible API for running local models.
    This client is optimized for local deployment with:
    - No authentication required (local)
    - Configurable base URL
    - No circuit breaker (local server expected to be stable)
    - Longer timeouts for large models
    """

    PROVIDER_NAME = "lmstudio"
    DEFAULT_BASE_URL = "http://localhost:1234/v1"
    DEFAULT_MODEL = "local-model"  # LM Studio uses the loaded model

    def __init__(
        self,
        api_key: str | None = None,  # Not required for local
        model: str | None = None,
        base_url: str | None = None,
        timeout: float = 300.0,  # Long timeout for local inference
        max_retries: int = 2,  # Fewer retries for local
        # Rate limiting
        rate_limit_per_minute: int | None = None,
    ):
        """
        Initialize LM Studio client.

        Args:
            api_key: Not required for local server (ignored)
            model: Model identifier (often ignored by LM Studio, uses loaded model)
            base_url: Local server URL (default: http://localhost:1234/v1)
            timeout: Request timeout in seconds (default longer for local models)
            max_retries: Max retry attempts (fewer for local)
            rate_limit_per_minute: Rate limit for requests per minute (None to disable)
        """
        import os

        # Allow overriding via environment variable
        base_url = base_url or os.environ.get("LMSTUDIO_BASE_URL", self.DEFAULT_BASE_URL)

        super().__init__(
            api_key=api_key or "not-required",  # Placeholder
            model=model or self.DEFAULT_MODEL,
            base_url=base_url,
            timeout=timeout,
            max_retries=max_retries,
            rate_limit_per_minute=rate_limit_per_minute,
        )

        self._client: httpx.AsyncClient | None = None

    async def _get_client(self) -> httpx.AsyncClient:
        """Get or create the HTTP client."""
        if self._client is None or self._client.is_closed:
            headers = {"Content-Type": "application/json"}

            # Add auth header if provided (some local servers may require it)
            if self.api_key and self.api_key != "not-required":
                headers["Authorization"] = f"Bearer {self.api_key}"

            self._client = httpx.AsyncClient(
                base_url=self.base_url,
                headers=headers,
                timeout=httpx.Timeout(self.timeout),
            )
        return self._client

    async def check_health(self) -> bool:
        """
        Check if LM Studio server is running.

        Returns:
            True if server is accessible, False otherwise
        """
        try:
            client = await self._get_client()
            response = await client.get("/models")
            return response.status_code == 200
        except Exception:
            return False

    async def list_models(self) -> list[dict]:
        """
        List available models on the LM Studio server.

        Returns:
            List of model information dicts
        """
        try:
            client = await self._get_client()
            response = await client.get("/models")
            if response.status_code == 200:
                data = response.json()
                return data.get("data", [])
            return []
        except Exception as e:
            logger.warning(f"Failed to list models: {e}")
            return []

    def _handle_error_response(self, response: httpx.Response) -> None:
        """Handle error responses from LM Studio server."""
        status_code = response.status_code

        try:
            error_data = response.json()
            error_message = error_data.get("error", {}).get("message", response.text)
        except Exception:
            error_message = response.text

        if status_code >= 500:
            raise LLMServerError(self.PROVIDER_NAME, status_code, error_message)
        else:
            raise LLMClientError(error_message, self.PROVIDER_NAME, status_code=status_code)

    async def generate(
        self,
        *,
        messages: list[dict] | None = None,
        prompt: str | None = None,
        temperature: float = 0.7,
        max_tokens: int | None = None,
        tools: list[dict] | None = None,
        stream: bool = False,
        stop: list[str] | None = None,
        **kwargs: Any,
    ) -> LLMResponse | AsyncIterator[str]:
        """
        Generate a response from LM Studio local model.

        Args:
            messages: Chat messages in OpenAI format
            prompt: Simple string prompt
            temperature: Sampling temperature
            max_tokens: Maximum tokens to generate
            tools: Tool definitions (limited support in local models)
            stream: If True, returns AsyncIterator
            stop: Stop sequences
            **kwargs: Additional parameters

        Returns:
            LLMResponse or AsyncIterator[str] for streaming
        """
        # Apply rate limiting before proceeding
        await self._apply_rate_limit()

        if stream:
            return self._generate_stream(
                messages=messages,
                prompt=prompt,
                temperature=temperature,
                max_tokens=max_tokens,
                tools=tools,
                stop=stop,
                **kwargs,
            )
        else:
            return await self._generate_non_stream(
                messages=messages,
                prompt=prompt,
                temperature=temperature,
                max_tokens=max_tokens,
                tools=tools,
                stop=stop,
                **kwargs,
            )

    async def _generate_non_stream(
        self,
        *,
        messages: list[dict] | None = None,
        prompt: str | None = None,
        temperature: float = 0.7,
        max_tokens: int | None = None,
        tools: list[dict] | None = None,
        stop: list[str] | None = None,
        **kwargs: Any,
    ) -> LLMResponse:
        """Non-streaming generation."""
        client = await self._get_client()

        # Build request payload (OpenAI-compatible)
        payload = {
            "model": self.model,
            "messages": self._build_messages(messages, prompt),
            "temperature": temperature,
        }

        if max_tokens is not None:
            payload["max_tokens"] = max_tokens
        if stop:
            payload["stop"] = stop

        # Note: most local models don't support tools well
        if tools:
            logger.warning("Tool calling may not be fully supported by local models")
            payload["tools"] = tools

        # Add additional kwargs (e.g., top_p, repeat_penalty)
        for key in ["top_p", "top_k", "repeat_penalty", "presence_penalty", "frequency_penalty"]:
            if key in kwargs:
                payload[key] = kwargs[key]

        # Retry logic for local server
        last_error = None
        for attempt in range(self.max_retries):
            try:
                response = await client.post("/chat/completions", json=payload)

                if response.status_code != 200:
                    self._handle_error_response(response)

                # Parse response
                try:
                    data = response.json()
                    choice = data["choices"][0]
                    message = choice["message"]

                    usage = data.get("usage", {})
                    finish_reason = choice.get("finish_reason", "stop")

                    llm_response = LLMResponse(
                        text=message.get("content", ""),
                        usage=usage,
                        model=data.get("model", self.model),
                        raw_response=data,
                        finish_reason=finish_reason,
                    )

                    self._update_stats(llm_response)
                    return llm_response

                except (KeyError, json.JSONDecodeError) as e:
                    raise LLMResponseParseError(self.PROVIDER_NAME, response.text) from e

            except httpx.TimeoutException:
                last_error = LLMTimeoutError(self.PROVIDER_NAME, self.timeout)
                logger.warning(f"Attempt {attempt + 1} timed out, retrying...")
            except httpx.ConnectError:
                last_error = LLMConnectionError(self.PROVIDER_NAME, self.base_url)
                logger.warning(f"Attempt {attempt + 1} connection failed, retrying...")
            except LLMClientError:
                raise  # Don't retry client errors

        # All retries exhausted
        if last_error:
            raise last_error
        raise LLMConnectionError(self.PROVIDER_NAME, self.base_url)

    async def _generate_stream(
        self,
        *,
        messages: list[dict] | None = None,
        prompt: str | None = None,
        temperature: float = 0.7,
        max_tokens: int | None = None,
        tools: list[dict] | None = None,  # noqa: ARG002
        stop: list[str] | None = None,
        **kwargs: Any,
    ) -> AsyncIterator[str]:
        """Streaming generation."""
        client = await self._get_client()

        # Build request payload
        payload = {
            "model": self.model,
            "messages": self._build_messages(messages, prompt),
            "temperature": temperature,
            "stream": True,
        }

        if max_tokens is not None:
            payload["max_tokens"] = max_tokens
        if stop:
            payload["stop"] = stop

        for key in ["top_p", "top_k", "repeat_penalty"]:
            if key in kwargs:
                payload[key] = kwargs[key]

        async def stream_generator():
            try:
                async with client.stream("POST", "/chat/completions", json=payload) as response:
                    if response.status_code != 200:
                        await response.aread()
                        self._handle_error_response(response)

                    async for line in response.aiter_lines():
                        if line.startswith("data: "):
                            data_str = line[6:]
                            if data_str.strip() == "[DONE]":
                                break

                            try:
                                data = json.loads(data_str)
                                delta = data["choices"][0].get("delta", {})
                                content = delta.get("content", "")
                                if content:
                                    yield content
                            except (json.JSONDecodeError, KeyError):
                                continue

            except httpx.TimeoutException:
                raise LLMTimeoutError(self.PROVIDER_NAME, self.timeout)
            except httpx.ConnectError:
                raise LLMConnectionError(self.PROVIDER_NAME, self.base_url)
            except Exception as e:
                if isinstance(e, LLMClientError):
                    raise
                raise LLMStreamError(self.PROVIDER_NAME, str(e)) from e

        return stream_generator()

    async def close(self) -> None:
        """Close the HTTP client."""
        if self._client and not self._client.is_closed:
            await self._client.aclose()
            self._client = None