File size: 11,268 Bytes
af02d93
 
accec81
 
 
 
 
 
af02d93
 
 
 
 
 
 
 
 
 
 
 
 
 
 
accec81
 
 
 
 
 
 
71b8430
accec81
eda6854
accec81
eda6854
af02d93
71b8430
eba525c
 
 
 
af02d93
 
 
 
 
 
accec81
af02d93
 
 
 
 
 
d6def15
eda6854
af02d93
 
 
accec81
af02d93
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
accec81
af02d93
 
 
 
accec81
af02d93
 
accec81
 
af02d93
accec81
af02d93
 
 
 
 
eda6854
 
 
 
 
accec81
eba525c
 
 
 
 
 
 
 
 
 
accec81
eba525c
accec81
eba525c
 
 
 
 
 
 
 
8e47a66
eba525c
 
 
 
 
 
 
 
 
 
accec81
 
eba525c
accec81
af02d93
f7a796b
accec81
eda6854
 
 
 
 
accec81
eda6854
 
eba525c
 
 
eda6854
 
 
accec81
eda6854
 
af02d93
eda6854
accec81
eda6854
 
 
 
 
 
accec81
eda6854
 
 
 
 
 
 
accec81
 
eda6854
accec81
 
eda6854
 
accec81
 
eba525c
accec81
eda6854
 
 
accec81
eba525c
accec81
eda6854
accec81
a167ec5
 
 
accec81
 
af02d93
a167ec5
 
 
 
 
 
7d211df
a167ec5
 
accec81
a167ec5
 
 
 
 
 
 
af02d93
 
accec81
eba525c
accec81
 
eda6854
af02d93
 
eda6854
8e47a66
 
 
 
 
 
 
eda6854
 
8e47a66
 
 
 
 
 
 
5ab49bb
eda6854
af02d93
 
eba525c
af02d93
 
accec81
 
 
af02d93
accec81
 
 
 
eba525c
 
af02d93
 
accec81
 
 
af02d93
eba525c
 
 
 
 
8e47a66
 
 
 
 
 
 
 
 
eba525c
 
 
 
 
 
 
 
 
af02d93
 
 
eba525c
accec81
af02d93
eda6854
13045e2
eda6854
eba525c
eda6854
 
 
 
 
eba525c
eda6854
 
 
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
from __future__ import annotations

import os
import json
import time
import uuid
import asyncio
import logging
from typing import Any, AsyncGenerator
from contextlib import asynccontextmanager

from dotenv import load_dotenv
from fastapi import FastAPI, HTTPException, Request, Depends
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import StreamingResponse, JSONResponse
from pydantic import BaseModel
from gradio_client import Client

load_dotenv()

# ---------------------------------------------------------------------------
# Config
# ---------------------------------------------------------------------------

API_KEY = os.getenv("API_KEY", "")
HF_SPACE_URL = os.getenv("HF_SPACE_URL", "")
MODEL_ID = os.getenv("MODEL_ID", "")

DEFAULT_TEMP = float(os.getenv("DEFAULT_TEMPERATURE", "0.6"))
DEFAULT_TOP_P = float(os.getenv("DEFAULT_TOP_P", "0.95"))
DEFAULT_TOKENS = int(os.getenv("DEFAULT_MAX_TOKENS", "16000"))

REQUEST_TIMEOUT = int(os.getenv("REQUEST_TIMEOUT", "120"))
MAX_RETRIES = int(os.getenv("MAX_RETRIES", "3"))
RETRY_BASE_DELAY = float(os.getenv("RETRY_BASE_DELAY", "1.5"))

MAX_INPUT_TOKENS = 16000  # stała wartość

# przybliżone przeliczenie: 1 token ~ 4 znaki
AVG_CHARS_PER_TOKEN = 4

logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(message)s")
log = logging.getLogger(__name__)

# ---------------------------------------------------------------------------
# Gradio client (singleton)
# ---------------------------------------------------------------------------

_client: Client | None = None

async def get_client() -> Client:
    global _client
    if _client is None:
        log.info("Connecting to %s", HF_SPACE_URL)
        _client = await asyncio.to_thread(Client, HF_SPACE_URL)
        log.info("Connected.")
    return _client

# ---------------------------------------------------------------------------
# Schemas
# ---------------------------------------------------------------------------

class Message(BaseModel):
    role: str
    content: str | list[dict] = ""
    name: str | None = None

class ChatCompletionRequest(BaseModel):
    model: str = MODEL_ID
    messages: list[Message]
    temperature: float = DEFAULT_TEMP
    top_p: float = DEFAULT_TOP_P
    max_tokens: int = DEFAULT_TOKENS
    stream: bool = False
    frequency_penalty: float = 0
    presence_penalty: float = 0
    stop: str | list[str] | None = None
    seed: int | None = None
    user: str | None = None

# ---------------------------------------------------------------------------
# Auth
# ---------------------------------------------------------------------------

async def verify_key(request: Request) -> None:
    if not API_KEY:
        return
    auth = request.headers.get("Authorization", "")
    if not auth.startswith("Bearer ") or auth[7:] != API_KEY:
        raise HTTPException(status_code=401, detail="Invalid or missing API key")

# ---------------------------------------------------------------------------
# Lifespan
# ---------------------------------------------------------------------------

@asynccontextmanager
async def lifespan(app: FastAPI):
    log.info("Startup: connecting to Gradio client...")
    await get_client()
    yield
    log.info("Shutdown.")

# ---------------------------------------------------------------------------
# Utilities
# ---------------------------------------------------------------------------

def _content_str(m: Message) -> str:
    if isinstance(m.content, str):
        return m.content
    text_parts = []
    for p in m.content:
        if isinstance(p, dict) and p.get("type") == "text":
            text_parts.append(p.get("text", "").strip())
    return "".join(text_parts)

def _token_count(text: str) -> int:
    return max(1, len(text) // AVG_CHARS_PER_TOKEN)

def _condense_messages(messages: list[Message], max_tokens: int) -> str:
    system_msgs = [m for m in messages if m.role == "system"]
    user_assistant = [m for m in messages if m.role in ("user", "assistant")]

    condensed_parts = []
    for m in system_msgs:
        condensed_parts.append(_content_str(m))

    tokens_so_far = sum(_token_count(part) for part in condensed_parts)

    for m in user_assistant:
        text = _content_str(m)
        tcount = _token_count(text)
        if tokens_so_far + tcount > max_tokens:
            remaining_tokens = max_tokens - tokens_so_far
            if remaining_tokens <= 0:
                continue
            approx_chars = remaining_tokens * AVG_CHARS_PER_TOKEN
            text = text[-approx_chars:]
            tcount = _token_count(text)
        condensed_parts.append(text)
        tokens_so_far += tcount

    return "\n".join(condensed_parts)

def _build_prompt(messages: list[Message]) -> str:
    prompt = _condense_messages(messages, MAX_INPUT_TOKENS)
    log.info("Final prompt token count: ~%d", _token_count(prompt))
    return prompt

# ---------------------------------------------------------------------------
# Extraction
# ---------------------------------------------------------------------------

def _extract_text(result: Any) -> str:
    if isinstance(result, tuple):
        data = result
    elif hasattr(result, "data"):
        data = result.data
    else:
        data = [result]

    conversation = None
    for item in data:
        if isinstance(item, dict) and "value" in item and isinstance(item["value"], list):
            conversation = item["value"]
            break
        elif isinstance(item, list):
            conversation = item
            break

    if not conversation:
        raise ValueError("Cannot extract conversation from result")

    last = conversation[-1]

    if isinstance(last, dict):
        content = last.get("content", "")
    elif isinstance(last, (list, tuple)) and len(last) >= 2:
        content = last[1] or ""
    else:
        content = str(last)

    if isinstance(content, list):
        parts = []
        for block in content:
            if isinstance(block, dict) and block.get("type") == "text":
                parts.append(block.get("content", block.get("text", "")))
        return "".join(parts).strip()
    return str(content).strip()

# ---------------------------------------------------------------------------
# Retry wrapper
# ---------------------------------------------------------------------------

async def _call_with_retries(prompt: str, req: ChatCompletionRequest) -> str:
    last_error = None
    for attempt in range(1, MAX_RETRIES + 1):
        try:
            return await asyncio.wait_for(_call_falcon_once(prompt, req), timeout=REQUEST_TIMEOUT)
        except Exception as e:
            last_error = e
            if attempt == MAX_RETRIES:
                break
            delay = RETRY_BASE_DELAY ** attempt
            log.warning("Attempt %d failed: %s | retrying in %.2fs", attempt, str(e), delay)
            await asyncio.sleep(delay)
    raise last_error

# ---------------------------------------------------------------------------
# Falcon call with explicit api_name
# ---------------------------------------------------------------------------

async def _call_falcon_once(prompt: str, req: ChatCompletionRequest) -> str:
    client = await get_client()
    settings = {
        "model": req.model,
        "temperature": req.temperature,
        "max_new_tokens": req.max_tokens,
        "top_p": req.top_p,
    }

    # Reset chat session
    await asyncio.to_thread(client.predict, api_name="/new_chat")

    # Add message with explicit api_name and settings
    result = await asyncio.to_thread(
        client.predict,
        prompt,  # pierwszy argument
        settings_form_value=settings,
        api_name="/add_message",  # <-- tutaj musi być endpoint z View API
    )
    return _extract_text(result)

# ---------------------------------------------------------------------------
# Streaming
# ---------------------------------------------------------------------------

async def _stream_sse(text: str, req: ChatCompletionRequest) -> AsyncGenerator[str, None]:
    cid = f"chatcmpl-{uuid.uuid4().hex}"
    created = int(time.time())
    for i in range(0, len(text), 8):
        chunk = {
            "id": cid,
            "object": "chat.completion.chunk",
            "created": created,
            "model": req.model,
            "choices": [{"index": 0, "delta": {"content": text[i:i+8]}, "finish_reason": None}]
        }
        yield f"data: {json.dumps(chunk)}\n\n"
        await asyncio.sleep(0.01)
    final_chunk = {
        "id": cid,
        "object": "chat.completion.chunk",
        "created": created,
        "model": req.model,
        "choices": [{"index": 0, "delta": {}, "finish_reason": "stop"}]
    }
    yield f"data: {json.dumps(final_chunk)}\n\n"
    yield "data: [DONE]\n\n"

# ---------------------------------------------------------------------------
# OpenAI-compatible response
# ---------------------------------------------------------------------------

def _make_response(text: str, req: ChatCompletionRequest) -> dict:
    pt = sum(len(_content_str(m)) for m in req.messages) // 4
    ct = len(text) // 4
    return {
        "id": f"chatcmpl-{uuid.uuid4().hex}",
        "object": "chat.completion",
        "created": int(time.time()),
        "model": req.model,
        "choices": [{"index": 0, "message": {"role": "assistant", "content": text}, "finish_reason": "stop"}],
        "usage": {"prompt_tokens": pt, "completion_tokens": ct, "total_tokens": pt + ct},
    }

# ---------------------------------------------------------------------------
# Routes
# ---------------------------------------------------------------------------

app = FastAPI(title="Foc", version="5.0.0", lifespan=lifespan)
app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_credentials=True, allow_methods=["*"], allow_headers=["*"])

@app.get("/")
async def root():
    return {
        "service": "FOC API",
        "version": "5.0.0",
        "endpoints": {
            "health": "/health",
            "models": "/v1/models",
            "chat": "/v1/chat/completions"
        }
    }

@app.get("/health")
async def health():
    return {"status": "ok", "model": MODEL_ID, "space": HF_SPACE_URL}

@app.get("/v1/models")
async def list_models(_: None = Depends(verify_key)):
    return {"object": "list", "data": [{"id": MODEL_ID, "object": "model", "created": 1710000000, "owned_by": "tiiuae"}]}

@app.post("/v1/chat/completions")
async def chat_completions(req: ChatCompletionRequest, _: None = Depends(verify_key)):
    prompt = _build_prompt(req.messages)
    log.info("Request | model=%s temp=%.2f tokens=%d stream=%s", req.model, req.temperature, req.max_tokens, req.stream)

    try:
        text = await _call_with_retries(prompt, req)
    except Exception:
        log.exception("Falcon failed after retries")
        raise HTTPException(status_code=502, detail="Model temporarily unavailable. Please try again.")

    if req.stream:
        return StreamingResponse(
            _stream_sse(text, req),
            media_type="text/event-stream",
            headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no", "Connection": "keep-alive"},
        )

    return JSONResponse(content=_make_response(text, req))