File size: 15,314 Bytes
7adfd09
 
 
 
 
 
 
 
 
99f0108
7adfd09
 
 
 
fb6570d
7adfd09
fb6570d
7adfd09
 
99f0108
fb6570d
41ee25d
7adfd09
99f0108
7adfd09
28340f8
 
 
 
fb6570d
99f0108
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7adfd09
 
 
28340f8
7adfd09
99f0108
 
 
 
 
 
 
 
 
 
 
 
 
 
7adfd09
 
 
99f0108
 
41ee25d
 
7adfd09
 
99f0108
 
 
7adfd09
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
41ee25d
99f0108
 
 
 
 
 
 
 
7adfd09
 
 
 
 
 
99f0108
 
 
41ee25d
99f0108
7adfd09
 
 
 
 
 
 
 
 
99f0108
 
 
 
 
 
 
 
41ee25d
99f0108
fb6570d
 
99f0108
fb6570d
 
99f0108
7adfd09
fb6570d
99f0108
 
41ee25d
7adfd09
 
99f0108
fb6570d
7adfd09
99f0108
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
41ee25d
99f0108
41ee25d
99f0108
 
 
 
 
 
 
41ee25d
99f0108
fb6570d
 
28340f8
 
 
fb6570d
 
99f0108
 
41ee25d
 
99f0108
41ee25d
 
99f0108
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fb6570d
41ee25d
99f0108
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fb6570d
41ee25d
99f0108
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
28340f8
99f0108
 
 
 
 
28340f8
fb6570d
28340f8
7adfd09
 
 
99f0108
 
 
7adfd09
 
 
fb6570d
 
7adfd09
 
99f0108
 
 
7adfd09
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
41ee25d
 
7adfd09
 
 
 
 
28340f8
7adfd09
 
 
 
 
 
28340f8
7adfd09
 
 
 
 
 
 
28340f8
7adfd09
 
 
 
99f0108
7adfd09
 
 
fb6570d
7adfd09
 
 
99f0108
 
7adfd09
 
 
99f0108
 
 
 
 
 
 
 
 
 
 
 
7adfd09
41ee25d
7adfd09
99f0108
 
 
28340f8
99f0108
41ee25d
fdf82b0
7adfd09
 
99f0108
 
 
 
 
 
 
 
7adfd09
99f0108
7adfd09
 
 
99f0108
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7adfd09
 
 
99f0108
 
7adfd09
 
 
 
 
 
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
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
from __future__ import annotations

import asyncio
import json
import logging
import os
import time
import uuid
from contextlib import asynccontextmanager
from typing import Dict, List, Optional, Union, Any

from fastapi import FastAPI, HTTPException, Request
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import JSONResponse, StreamingResponse
from huggingface_hub import hf_hub_download
from pydantic import BaseModel, Field, ValidationError
from llama_cpp import Llama

# ---------- Configuration ----------
DEFAULT_MODEL_NAME = os.getenv("DEFAULT_MODEL_NAME", "bonsai-1.7b")
LOCAL_MODEL_DIR = os.getenv("LOCAL_MODEL_DIR", "/data/models")
MAX_NEW_TOKENS_DEFAULT = int(os.getenv("MAX_NEW_TOKENS_DEFAULT", "256"))
API_KEY = os.getenv("API_KEY", None)
HF_TOKEN = os.getenv("HF_TOKEN")

# Performance settings
N_CTX = int(os.getenv("N_CTX", "4096"))
N_THREADS = int(os.getenv("N_THREADS", "4"))
N_BATCH = int(os.getenv("N_BATCH", "512"))

# ---------- Model Registry ----------
MODEL_REGISTRY: Dict[str, Dict[str, str]] = {
    "bonsai-1.7b": {
        "repo_id": "lilyanatia/Bonsai-1.7B-requantized",
        "filename": "Bonsai-1.7B-IQ1_S.gguf",
    },
    "bonsai-4b": {
        "repo_id": "lilyanatia/Bonsai-4B-requantized",
        "filename": "Bonsai-4B-IQ1_S.gguf",
    },
    "bonsai-8b": {
        "repo_id": "lilyanatia/Bonsai-8B-requantized",
        "filename": "Bonsai-8B-IQ1_S.gguf",
    },
}

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("uvicorn.error")

# ---------- Pydantic Models ----------
class Message(BaseModel):
    role: str = Field(..., pattern="^(system|user|assistant|tool)$")
    content: Optional[str] = None
    tool_calls: Optional[List[Dict[str, Any]]] = None
    tool_call_id: Optional[str] = None
    name: Optional[str] = None

class ToolFunction(BaseModel):
    name: str
    description: Optional[str] = None
    parameters: Optional[Dict[str, Any]] = None

class Tool(BaseModel):
    type: str = "function"
    function: ToolFunction

class ChatCompletionRequest(BaseModel):
    messages: List[Message]
    model: str = Field(default=DEFAULT_MODEL_NAME)
    max_tokens: int = Field(default=MAX_NEW_TOKENS_DEFAULT, ge=1, le=2048)
    temperature: float = Field(default=0.7, ge=0.0, le=2.0)
    top_p: float = Field(default=0.95, gt=0.0, le=1.0)
    stream: bool = False
    stop: Optional[Union[str, List[str]]] = None
    tools: Optional[List[Tool]] = None
    tool_choice: Optional[Union[str, Dict[str, Any]]] = None
    response_format: Optional[Dict[str, str]] = None

class ChatCompletionResponseChoice(BaseModel):
    index: int
    message: Message
    finish_reason: str

class Usage(BaseModel):
    prompt_tokens: int
    completion_tokens: int
    total_tokens: int

class ChatCompletionResponse(BaseModel):
    id: str
    object: str = "chat.completion"
    created: int
    model: str
    choices: List[ChatCompletionResponseChoice]
    usage: Usage

class ModelInfo(BaseModel):
    id: str
    object: str = "model"
    created: int
    owned_by: str = "lilyanatia"

class ModelListResponse(BaseModel):
    object: str = "list"
    data: List[ModelInfo]

class ErrorResponse(BaseModel):
    error: str
    detail: Optional[str] = None

# ---------- Global State ----------
current_model_name: Optional[str] = None
llm: Optional[Llama] = None
model_load_error: Optional[str] = None
MODEL_LOCK = asyncio.Lock()
DOWNLOADED_MODELS = set()

# ---------- Helper Functions ----------
def _verify_api_key(request: Request) -> None:
    if API_KEY is None:
        return
    auth = request.headers.get("X-API-Key")
    if not auth or auth != API_KEY:
        raise HTTPException(status_code=401, detail="Invalid or missing API key")

def _download_model(model_name: str) -> str:
    """Downloads a model if it's not already present."""
    if model_name not in MODEL_REGISTRY:
        raise HTTPException(status_code=400, detail=f"Model '{model_name}' not found in registry.")
    
    model_info = MODEL_REGISTRY[model_name]
    repo_id = model_info["repo_id"]
    filename = model_info["filename"]
    os.makedirs(LOCAL_MODEL_DIR, exist_ok=True)
    local_path = os.path.join(LOCAL_MODEL_DIR, filename)

    if os.path.exists(local_path):
        logger.info(f"Model '{model_name}' already downloaded at {local_path}")
        return local_path

    logger.info(f"Downloading model '{model_name}' from {repo_id}/{filename}...")
    try:
        hf_hub_download(
            repo_id=repo_id,
            filename=filename,
            local_dir=LOCAL_MODEL_DIR,
            token=HF_TOKEN,
        )
        logger.info(f"Model '{model_name}' downloaded successfully.")
        return local_path
    except Exception as e:
        logger.error(f"Model download failed for '{model_name}': {e}")
        raise HTTPException(status_code=500, detail=f"Failed to download model: {str(e)}")

async def _precache_all_models():
    """Downloads all models in the registry at startup."""
    logger.info("Pre-caching all models in registry...")
    download_tasks = []
    for model_name in MODEL_REGISTRY.keys():
        download_tasks.append(asyncio.to_thread(_download_model, model_name))
    
    results = await asyncio.gather(*download_tasks, return_exceptions=True)
    for model_name, result in zip(MODEL_REGISTRY.keys(), results):
        if isinstance(result, Exception):
            logger.error(f"Failed to pre-cache model '{model_name}': {result}")
        else:
            DOWNLOADED_MODELS.add(model_name)
            logger.info(f"Model '{model_name}' is ready.")
    
    logger.info(f"Pre-caching complete. {len(DOWNLOADED_MODELS)}/{len(MODEL_REGISTRY)} models cached.")

async def _ensure_model_loaded(model_name: str):
    """Loads the specified model, downloading it first if necessary."""
    global llm, current_model_name, model_load_error
    async with MODEL_LOCK:
        if current_model_name == model_name and llm is not None:
            return

        if llm is not None:
            logger.info(f"Unloading previous model '{current_model_name}'...")
            del llm
            llm = None
            current_model_name = None

        try:
            model_path = _download_model(model_name)
            llm = Llama(
                model_path=model_path,
                n_ctx=N_CTX,
                n_threads=N_THREADS,
                n_batch=N_BATCH,
                verbose=False,
            )
            current_model_name = model_name
            logger.info(f"Model '{model_name}' loaded successfully.")
        except Exception as e:
            model_load_error = str(e)
            logger.exception(f"Model loading failed for '{model_name}'")
            raise HTTPException(status_code=503, detail=f"Model unavailable: {model_load_error}")

def _build_chat_prompt(messages: List[Message]) -> List[Dict[str, Any]]:
    """Convert Pydantic messages to dict format for llama.cpp."""
    formatted = []
    for msg in messages:
        msg_dict = {"role": msg.role, "content": msg.content}
        if msg.tool_calls:
            msg_dict["tool_calls"] = msg.tool_calls
        if msg.tool_call_id:
            msg_dict["tool_call_id"] = msg.tool_call_id
        if msg.name:
            msg_dict["name"] = msg.name
        formatted.append(msg_dict)
    return formatted

def _convert_tools(tools: Optional[List[Tool]]) -> Optional[List[Dict[str, Any]]]:
    """Convert Pydantic tools to dict format for llama.cpp."""
    if not tools:
        return None
    return [tool.model_dump() for tool in tools]

async def _generate_full(
    prompt: List[Dict[str, Any]],
    max_new_tokens: int,
    temperature: float,
    top_p: float,
    stop_sequences: Optional[List[str]] = None,
    tools: Optional[List[Dict[str, Any]]] = None,
    tool_choice: Optional[Union[str, Dict[str, Any]]] = None,
    response_format: Optional[Dict[str, str]] = None,
) -> Dict[str, Any]:
    if llm is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    kwargs = {
        "messages": prompt,
        "max_tokens": max_new_tokens,
        "temperature": temperature,
        "top_p": top_p,
        "stop": stop_sequences,
        "stream": False,
    }
    if tools:
        kwargs["tools"] = tools
    if tool_choice:
        kwargs["tool_choice"] = tool_choice
    if response_format:
        kwargs["response_format"] = response_format
    
    result = await asyncio.to_thread(lambda: llm.create_chat_completion(**kwargs))
    return result

async def _generate_stream(
    prompt: List[Dict[str, Any]],
    max_new_tokens: int,
    temperature: float,
    top_p: float,
    stop_sequences: Optional[List[str]] = None,
    tools: Optional[List[Dict[str, Any]]] = None,
    tool_choice: Optional[Union[str, Dict[str, Any]]] = None,
    response_format: Optional[Dict[str, str]] = None,
):
    if llm is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    kwargs = {
        "messages": prompt,
        "max_tokens": max_new_tokens,
        "temperature": temperature,
        "top_p": top_p,
        "stop": stop_sequences,
        "stream": True,
    }
    if tools:
        kwargs["tools"] = tools
    if tool_choice:
        kwargs["tool_choice"] = tool_choice
    if response_format:
        kwargs["response_format"] = response_format
    
    def sync_gen():
        for chunk in llm.create_chat_completion(**kwargs):
            yield chunk
    
    for chunk in await asyncio.to_thread(list, sync_gen()):
        yield chunk
        await asyncio.sleep(0)

# ---------- FastAPI App ----------
@asynccontextmanager
async def lifespan(app: FastAPI):
    try:
        await _precache_all_models()
        await _ensure_model_loaded(DEFAULT_MODEL_NAME)
        logger.info(f"Default model '{DEFAULT_MODEL_NAME}' loaded successfully")
    except Exception as e:
        logger.error(f"Startup model load failed: {e}")
    yield
    global llm
    llm = None

app = FastAPI(
    title="Bonsai Multi-Model Inference API",
    version="3.0.0",
    description="Lightning-fast inference for Bonsai LLMs with tool calling support.",
    docs_url="/docs",
    redoc_url="/redoc",
    lifespan=lifespan,
)

app.add_middleware(
    CORSMiddleware,
    allow_origins=os.getenv("ALLOW_ORIGINS", "*").split(","),
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

@app.middleware("http")
async def auth_middleware(request: Request, call_next):
    _verify_api_key(request)
    response = await call_next(request)
    return response

@app.exception_handler(HTTPException)
async def http_exception_handler(request, exc):
    return JSONResponse(
        status_code=exc.status_code,
        content=ErrorResponse(error=exc.detail, detail=str(exc.detail)).model_dump(),
    )

@app.exception_handler(ValidationError)
async def validation_exception_handler(request, exc):
    return JSONResponse(
        status_code=422,
        content=ErrorResponse(error="Validation error", detail=str(exc)).model_dump(),
    )

@app.exception_handler(Exception)
async def generic_exception_handler(request, exc):
    logger.exception("Unhandled exception")
    return JSONResponse(
        status_code=500,
        content=ErrorResponse(error="Internal server error", detail=str(exc)).model_dump(),
    )

@app.get("/", summary="Root")
def root():
    return {"message": "Bonsai Multi-Model API is running", "docs": "/docs"}

@app.get("/health", summary="Health check")
def health():
    loaded = llm is not None
    return {
        "status": "ok" if loaded else "degraded",
        "model_loaded": loaded,
        "current_model": current_model_name,
        "cached_models": list(DOWNLOADED_MODELS),
        "error": model_load_error if model_load_error else None,
    }

@app.get("/v1/models", response_model=ModelListResponse, summary="List available models")
def list_models():
    models = []
    for name in MODEL_REGISTRY.keys():
        models.append(ModelInfo(id=name, created=int(time.time())))
    return ModelListResponse(data=models)

@app.get("/v1/models/{model_name}", response_model=ModelInfo, summary="Get model information")
def get_model(model_name: str):
    if model_name not in MODEL_REGISTRY:
        raise HTTPException(status_code=404, detail=f"Model '{model_name}' not found")
    return ModelInfo(id=model_name, created=int(time.time()))

@app.post("/v1/chat/completions", response_model=ChatCompletionResponse)
async def chat_completions(req: ChatCompletionRequest):
    model_name = req.model or DEFAULT_MODEL_NAME
    await _ensure_model_loaded(model_name)
    
    prompt = _build_chat_prompt(req.messages)
    tools = _convert_tools(req.tools)
    stop_seq = req.stop if isinstance(req.stop, list) else ([req.stop] if req.stop else None)

    if req.stream:
        async def stream_generator():
            yield f"data: {json.dumps({'id': f'chatcmpl-{uuid.uuid4().hex[:12]}', 'object': 'chat.completion.chunk', 'created': int(time.time()), 'model': model_name, 'choices': [{'index': 0, 'delta': {'role': 'assistant'}, 'finish_reason': None}]})}\n\n"
            async for chunk in _generate_stream(prompt, req.max_tokens, req.temperature, req.top_p, stop_seq, tools, req.tool_choice, req.response_format):
                delta = {}
                if "choices" in chunk and len(chunk["choices"]) > 0:
                    choice = chunk["choices"][0]
                    if "delta" in choice:
                        delta = choice["delta"]
                yield f"data: {json.dumps({'id': f'chatcmpl-{uuid.uuid4().hex[:12]}', 'object': 'chat.completion.chunk', 'created': int(time.time()), 'model': model_name, 'choices': [{'index': 0, 'delta': delta, 'finish_reason': None}]})}\n\n"
                await asyncio.sleep(0)
            yield f"data: {json.dumps({'id': f'chatcmpl-{uuid.uuid4().hex[:12]}', 'object': 'chat.completion.chunk', 'created': int(time.time()), 'model': model_name, 'choices': [{'index': 0, 'delta': {}, 'finish_reason': 'stop'}]})}\n\n"
            yield "data: [DONE]\n\n"
        return StreamingResponse(stream_generator(), media_type="text/event-stream")
    else:
        result = await _generate_full(prompt, req.max_tokens, req.temperature, req.top_p, stop_seq, tools, req.tool_choice, req.response_format)
        choice = result["choices"][0]
        message_data = choice.get("message", {})
        assistant_msg = Message(
            role=message_data.get("role", "assistant"),
            content=message_data.get("content"),
            tool_calls=message_data.get("tool_calls"),
        )
        finish_reason = choice.get("finish_reason", "stop")
        usage_data = result.get("usage", {})
        usage = Usage(
            prompt_tokens=usage_data.get("prompt_tokens", 0),
            completion_tokens=usage_data.get("completion_tokens", 0),
            total_tokens=usage_data.get("total_tokens", 0),
        )
        return ChatCompletionResponse(
            id=f"chatcmpl-{uuid.uuid4().hex[:12]}",
            created=int(time.time()),
            model=model_name,
            choices=[ChatCompletionResponseChoice(index=0, message=assistant_msg, finish_reason=finish_reason)],
            usage=usage,
        )

if __name__ == "__main__":
    import uvicorn
    uvicorn.run("app:app", host="0.0.0.0", port=7860, reload=False)