Blizzard
84d1a1dd3a
mcp-go 接通向量 RAG：embedding(OpenAI 兼容 provider 抽象) + Milvus 真实连接， kb_ingest 入库、wiki_search 真检索。retriever 节点一行不改即从桩变真。 - mcp-go internal/rag: embed.go(OpenAI 兼容 /embeddings 客户端) + milvus.go(milvus-sdk-go 真连，集合按首次 embedding 维度懒建+AUTOINDEX/COSINE索引+加载，insert/向量search) + rag.go(Engine: 切块→embed→insert / embed query→search；embedding 或 Milvus 缺则降级) - mcp-go gateway: 新工具 kb_ingest，wiki_search 换真(RAG 向量检索，kb 过滤 topK) - mcp-go main: rag.Open 读 MILVUS_ADDR/EMBED_BASE_URL/EMBED_API_KEY/EMBED_MODEL 环境变量 - gateway: POST /api/v1/kb/ingest → kb_ingest（供知识库页/脚本） - scripts/mock_embeddings.py: 确定性词法向量(字+bigram 哈希)，无真 key 验证检索 - 开发期 embedding 接在线 API(无真 key 用 mock)，见 llm-provider-strategy - 验证: 全模块 build✓ + e2e PASS; live——入库5条→Milvus；retriever 节点查'向量数据库' →召回 Milvus 那条→DeepSeek 答'Milvus'；查'知识图谱'→Neo4j（向量检索区分正确） 注: 当前向量单路；Bleve/Neo4j 融合 + rerank + 真实语义 embedding 为后续。 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
66 lines
2.4 KiB
Python
66 lines
2.4 KiB
Python
#!/usr/bin/env python3
|
|
"""最小 OpenAI 兼容 embeddings mock —— 无真实 embedding key 时验证 RAG 链路。
|
|
确定性"词法向量"：把文本的字 + 字bigram 哈希进固定维向量并归一化，
|
|
词面重叠越多余弦越高（足以演示"查询召回相关文档"）。真实语义需插真实 embedding API。
|
|
GET /models → 200
|
|
POST /embeddings → {data:[{embedding:[...]}]}
|
|
用法: python3 scripts/mock_embeddings.py 11888
|
|
"""
|
|
import hashlib
|
|
import json
|
|
import math
|
|
import sys
|
|
from http.server import BaseHTTPRequestHandler, HTTPServer
|
|
|
|
DIM = 256
|
|
|
|
|
|
def embed(text: str):
|
|
vec = [0.0] * DIM
|
|
t = text.strip()
|
|
toks = list(t) + [t[i:i + 2] for i in range(len(t) - 1)] # 字 + 字bigram
|
|
for tok in toks:
|
|
h = int(hashlib.md5(tok.encode("utf-8")).hexdigest(), 16)
|
|
vec[h % DIM] += 1.0
|
|
norm = math.sqrt(sum(x * x for x in vec)) or 1.0
|
|
return [x / norm for x in vec]
|
|
|
|
|
|
class H(BaseHTTPRequestHandler):
|
|
def log_message(self, *a):
|
|
pass
|
|
|
|
def do_GET(self):
|
|
if self.path.endswith("/models"):
|
|
body = json.dumps({"object": "list", "data": [{"id": "mock-embed"}]}).encode()
|
|
self.send_response(200)
|
|
self.send_header("Content-Type", "application/json")
|
|
self.send_header("Content-Length", str(len(body)))
|
|
self.end_headers()
|
|
self.wfile.write(body)
|
|
else:
|
|
self.send_response(404)
|
|
self.end_headers()
|
|
|
|
def do_POST(self):
|
|
n = int(self.headers.get("Content-Length", 0))
|
|
req = json.loads(self.rfile.read(n) or b"{}")
|
|
inp = req.get("input", [])
|
|
if isinstance(inp, str):
|
|
inp = [inp]
|
|
data = [{"object": "embedding", "index": i, "embedding": embed(t)} for i, t in enumerate(inp)]
|
|
body = json.dumps({"object": "list", "data": data, "model": req.get("model", "mock-embed")}).encode()
|
|
sys.stderr.write(f"[mock-embed] /embeddings n={len(inp)} dim={DIM}\n")
|
|
sys.stderr.flush()
|
|
self.send_response(200)
|
|
self.send_header("Content-Type", "application/json")
|
|
self.send_header("Content-Length", str(len(body)))
|
|
self.end_headers()
|
|
self.wfile.write(body)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
port = int(sys.argv[1]) if len(sys.argv) > 1 else 11888
|
|
print(f"[mock-embed] listening on :{port} (dim={DIM})")
|
|
HTTPServer(("127.0.0.1", port), H).serve_forever()
|