sanic server script

Signed-off-by: guangli.bao <[email protected]>

Author: tukwila

pyproject.toml

@@ -143,7 +143,7 @@ exclude = ["venv", ".tox"]
 follow_imports = 'silent'
 
 [[tool.mypy.overrides]]
-module = ["datasets.*", "transformers.*", "setuptools.*", "setuptools_git_versioning.*"]
+module = ["datasets.*", "transformers.*", "setuptools.*", "setuptools_git_versioning.*", "sanic.*"]
 ignore_missing_imports=true
 
 

tests/unit/sanic_server.py

@@ -0,0 +1,242 @@
+#!/usr/bin/env python3
+"""
+sanic_server.py
+
+Simple Sanic-based mock server that implements common OpenAI / vLLM-compatible routes:
+- GET  /                    : health
+- GET  /v1/models           : list models
+- POST /v1/chat/completions : chat completions (supports streaming via ?stream=true)
+- POST /v1/completions      : classic completions
+- POST /v1/embeddings       : fake embeddings
+- POST /v1/moderations      : fake moderation
+
+Usage:
+    pip install sanic==25.3.0 or latest
+Command:
+    python sanic_server.py or \
+    python sanic_server.py --host=0.0.0.0 --port=8000 --workers=1 --debug
+"""
+
+import argparse
+import asyncio
+import json
+import random
+
+from sanic import Sanic
+from sanic.request import Request
+from sanic.response import ResponseStream
+from sanic.response import json as sjson
+
+app = Sanic("sanic_server")
+
+
+# ---------- utils ----------
+
+
+def fake_tokenize(text: str) -> list[str]:
+    # crude whitespace tokenizer for token counting
+    return text.strip().split()
+
+
+def make_choice_text(prompt: str) -> str:
+    # Very simple deterministic reply generator
+    # Echo some truncated summary for testing
+    tail = prompt.strip()[:120]
+    return f"Mock reply summarizing: {tail}"
+
+
+def now_ms() -> int:
+    return int(asyncio.get_event_loop().time() * 1000)
+
+
+# ---------- routes ----------
+
+
+@app.get("/")
+async def health(request: Request):
+    return sjson({"ok": True, "msg": "mock openai/vllm server"})
+
+
+@app.get("/v1/models")
+async def list_models(request: Request):
+    # minimal model list
+    models = [
+        {"id": "mock-qwen-2.5", "object": "model", "owned_by": "mock"},
+        {"id": "facebook/opt-125m", "object": "model", "owned_by": "mock"},
+    ]
+    return sjson({"object": "list", "data": models})
+
+
+@app.post("/v1/chat/completions")
+async def chat_completions(request: Request):
+    """
+    Supports:
+      - JSON body with 'messages' (OpenAI format)
+      - query param stream=true or JSON {'stream': true}
+          => responds with text/event-stream chunks containing 'data: {json}\n\n'
+    """
+    body = request.json or {}
+    stream_mode = False
+    if request.args.get("stream", "false").lower() == "true":
+        stream_mode = True
+
+    messages = body.get("messages", [])
+    prompt_text = ""
+    if isinstance(messages, list) and messages:
+        # approximate prompt as concatenation of last user message(s)
+        for m in messages:
+            role = m.get("role", "")
+            content = m.get("content", "")
+            if role == "user":
+                prompt_text += content + " "
+
+    # build a deterministic reply
+    reply = make_choice_text(prompt_text or "hello")
+    prompt_tokens = len(fake_tokenize(prompt_text))
+    completion_tokens = len(fake_tokenize(reply))
+
+    # create response object (non-streaming)
+    def make_response_obj():
+        return {
+            "id": f"cmpl-mock-{random.randint(1000, 9999)}",
+            "object": "chat.completion",
+            "created": now_ms(),
+            "model": body.get("model", "mock-qwen-2.5"),
+            "usage": {
+                "prompt_tokens": prompt_tokens,
+                "completion_tokens": completion_tokens,
+                "total_tokens": prompt_tokens + completion_tokens,
+            },
+            "choices": [
+                {
+                    "index": 0,
+                    "message": {"role": "assistant", "content": reply},
+                    "finish_reason": "stop",
+                }
+            ],
+        }
+
+    if not stream_mode:
+        return sjson(make_response_obj())
+
+    # streaming mode: SSE-style chunks with 'data: <json>\n\n'
+    async def streaming_fn(resp):
+        # send an initial "response.start" like chunk
+        await resp.write(
+            f"data: \
+                    {json.dumps({'type': 'response.start', 'created': now_ms()})}\n\n"
+        )
+
+        # simulate token-by-token streaming
+        tokens = fake_tokenize(reply)
+        chunk_text = ""
+        for i, tk in enumerate(tokens):
+            chunk_text += tk + (" " if i < len(tokens) - 1 else "")
+            chunk_payload = {
+                "id": f"cmpl-mock-{random.randint(1000, 9999)}",
+                "object": "chat.completion.chunk",
+                "created": now_ms(),
+                "model": body.get("model", "mock-qwen-2.5"),
+                "choices": [
+                    {
+                        "delta": {"content": tk + (" " if i < len(tokens) - 1 else "")},
+                        "index": 0,
+                        "finish_reason": None,
+                    }
+                ],
+            }
+            # write chunk
+            await resp.write(f"data: {json.dumps(chunk_payload)}\n\n")
+            # small jitter between tokens
+            await asyncio.sleep(0.03)
+        # final done event
+        done_payload = {"type": "response.done", "created": now_ms()}
+        await resp.write(f"data: {json.dumps(done_payload)}\n\n")
+
+    headers = {"Content-Type": "text/event-stream", "Cache-Control": "no-cache"}
+    return ResponseStream(streaming_fn, headers=headers)
+
+
+@app.post("/v1/completions")
+async def completions(request: Request):
+    body = request.json or {}
+    prompt = body.get("prompt") or (
+        body.get("messages")
+        and " ".join([m.get("content", "") for m in body.get("messages", [])])
+    )
+    if not prompt:
+        prompt = "hello"
+    # optional max_tokens
+    max_tokens = int(body.get("max_tokens", 64))
+    reply = make_choice_text(prompt)
+    tokenized = fake_tokenize(reply)[:max_tokens]
+    text_out = " ".join(tokenized)
+
+    prompt_tokens = len(fake_tokenize(prompt))
+    completion_tokens = len(tokenized)
+
+    resp = {
+        "id": f"cmpl-mock-{random.randint(1000, 9999)}",
+        "object": "text_completion",
+        "created": now_ms(),
+        "model": body.get("model", "mock-qwen-2.5"),
+        "choices": [{"text": text_out, "index": 0, "finish_reason": "stop"}],
+        "usage": {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": prompt_tokens + completion_tokens,
+        },
+    }
+    # simulate a small server-side latency
+    await asyncio.sleep(0.01)
+    return sjson(resp)
+
+
+@app.post("/v1/embeddings")
+async def embeddings(request: Request):
+    body = request.json or {}
+    inputs = body.get("input") or body.get("inputs") or []
+    if isinstance(inputs, str):
+        inputs = [inputs]
+    # produce deterministic embedding length 16
+    dim = int(request.args.get("dim", body.get("dim", 16)))
+    out = []
+    for i, txt in enumerate(inputs):
+        # make pseudo-random but deterministic numbers based on hash
+        seed = abs(hash(txt)) % (10**8)
+        random.seed(seed)
+        vec = [round((random.random() - 0.5), 6) for _ in range(dim)]
+        out.append({"object": "embedding", "embedding": vec, "index": i})
+    return sjson({"data": out, "model": body.get("model", "mock-embed-1")})
+
+
+@app.post("/v1/moderations")
+async def moderations(request: Request):
+    body = request.json or {}
+    input_text = body.get("input") or ""
+    # super naive: classify as 'flagged' if contains "bad"
+    flagged = "bad" in input_text.lower()
+    return sjson(
+        {
+            "id": "mod-mock-1",
+            "model": body.get("model", "mock-moderation"),
+            "results": [{"flagged": flagged}],
+        }
+    )
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(prog="sanic_server")
+    parser.add_argument("--host", default="127.0.0.1")
+    parser.add_argument("--port", default=8000, type=int)
+    parser.add_argument("--debug", action="store_true")
+    parser.add_argument("--workers", default=1, type=int)
+    args = parser.parse_args()
+
+    app.run(
+        host=args.host,
+        port=args.port,
+        debug=args.debug,
+        workers=args.workers,
+        access_log=False,
+    )