Refactor setup: new/update detection, split Python servers, add MCP server
- laptop_full_setup.sh now handles both fresh install and updates cleanly (detects existing install, --force flag to overwrite config files) - server.py: standalone RAG server with AST-aware code chunking (Python), pattern-split for JS/TS/Go, /ingest/repo and /webhook/gitea|github endpoints - mcp_server.py: new MCP server (port 8002/SSE) with Claude Code-equivalent tools: bash, file ops, ripgrep search, git ops, Gitea API, GitHub API, RAG repo ingest - docker-compose: adds mcp-server service, workspace volume, env_file for tokens - .env preserved on update (tokens never overwritten) - GPU: OLLAMA_NUM_GPU=999 auto-adapts to any VRAM size (no hard-coded 6GB) - ZIM downloads remain in kiwix_download.sh (separate, large files) https://claude.ai/code/session_012gDnantBmFTWZGCiKyjazx
This commit is contained in:
Claude
Regular → Executable
+326
-545
File diff suppressed because it is too large
Load Diff
+203
@@ -0,0 +1,203 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
MCP Server — Claude Code-equivalent tools for Open WebUI / Claude Code CLI.
|
||||
Tools: bash, file read/write/list, code search, git ops, Gitea API, repo ingest.
|
||||
Connects via SSE on port 8002 — add to Open WebUI Tools or ~/.claude/mcp.json
|
||||
"""
|
||||
import os, subprocess, textwrap
|
||||
from pathlib import Path
|
||||
|
||||
import httpx
|
||||
from mcp.server.fastmcp import FastMCP
|
||||
|
||||
WORKSPACE = Path(os.getenv("WORKSPACE_DIR", "/workspace"))
|
||||
REPOS_DIR = Path(os.getenv("REPOS_DIR", "/repos"))
|
||||
GITEA_URL = os.getenv("GITEA_URL", "http://gitea:3000")
|
||||
GITEA_TOKEN = os.getenv("GITEA_TOKEN", "")
|
||||
GITHUB_TOKEN= os.getenv("GITHUB_TOKEN","")
|
||||
RAG_URL = os.getenv("RAG_URL", "http://rag-server:8001")
|
||||
|
||||
mcp = FastMCP("local-dev-tools")
|
||||
|
||||
# ── bash ──────────────────────────────────────────────────────────────────────
|
||||
@mcp.tool()
|
||||
def bash(command: str, cwd: str = "") -> str:
|
||||
"""Run a shell command. Default cwd is /workspace."""
|
||||
work = Path(cwd) if cwd else WORKSPACE
|
||||
work.mkdir(parents=True, exist_ok=True)
|
||||
try:
|
||||
r = subprocess.run(command, shell=True, cwd=work, timeout=120,
|
||||
capture_output=True, text=True)
|
||||
out = r.stdout + (f"\n[stderr]\n{r.stderr}" if r.stderr else "")
|
||||
if r.returncode != 0:
|
||||
out += f"\n[exit {r.returncode}]"
|
||||
return out or "(no output)"
|
||||
except subprocess.TimeoutExpired:
|
||||
return "[timeout after 120s]"
|
||||
except Exception as e:
|
||||
return f"[error] {e}"
|
||||
|
||||
# ── file ops ──────────────────────────────────────────────────────────────────
|
||||
@mcp.tool()
|
||||
def read_file(path: str) -> str:
|
||||
"""Read a file. Use absolute path or relative to /workspace."""
|
||||
p = Path(path) if Path(path).is_absolute() else WORKSPACE / path
|
||||
if not p.exists():
|
||||
return f"[not found] {p}"
|
||||
if p.stat().st_size > 500_000:
|
||||
return f"[too large — {p.stat().st_size//1024}KB]"
|
||||
return p.read_text(encoding="utf-8", errors="replace")
|
||||
|
||||
@mcp.tool()
|
||||
def write_file(path: str, content: str) -> str:
|
||||
"""Write content to a file (creates parent dirs). Relative to /workspace."""
|
||||
p = Path(path) if Path(path).is_absolute() else WORKSPACE / path
|
||||
p.parent.mkdir(parents=True, exist_ok=True)
|
||||
p.write_text(content, encoding="utf-8")
|
||||
return f"Wrote {len(content)} chars to {p}"
|
||||
|
||||
@mcp.tool()
|
||||
def list_files(path: str = "", pattern: str = "**/*") -> str:
|
||||
"""List files matching a glob pattern."""
|
||||
base = Path(path) if path else WORKSPACE
|
||||
if not base.exists():
|
||||
return f"[not found] {base}"
|
||||
files = sorted(str(f.relative_to(base)) for f in base.glob(pattern) if f.is_file())
|
||||
return "\n".join(files[:500]) or "(empty)"
|
||||
|
||||
@mcp.tool()
|
||||
def search_code(query: str, path: str = "", glob: str = "",
|
||||
case_sensitive: bool = False) -> str:
|
||||
"""Search file contents with ripgrep. Returns file:line matches."""
|
||||
base = path or str(WORKSPACE)
|
||||
cmd = ["rg", "--line-number", "--no-heading"]
|
||||
if not case_sensitive:
|
||||
cmd.append("-i")
|
||||
if glob:
|
||||
cmd += ["-g", glob]
|
||||
cmd += [query, base]
|
||||
try:
|
||||
r = subprocess.run(cmd, capture_output=True, text=True, timeout=30)
|
||||
lines = r.stdout.strip().splitlines()
|
||||
if len(lines) > 200:
|
||||
lines = lines[:200] + [f"… ({len(r.stdout.splitlines())-200} more)"]
|
||||
return "\n".join(lines) or "(no matches)"
|
||||
except FileNotFoundError:
|
||||
# ripgrep not installed, fall back to grep
|
||||
r = subprocess.run(["grep", "-rn", query, base],
|
||||
capture_output=True, text=True, timeout=30)
|
||||
return r.stdout[:8000] or "(no matches)"
|
||||
except Exception as e:
|
||||
return f"[error] {e}"
|
||||
|
||||
# ── git ───────────────────────────────────────────────────────────────────────
|
||||
def _git(args: list[str], repo: str = "") -> str:
|
||||
cwd = Path(repo) if repo else WORKSPACE
|
||||
r = subprocess.run(["git"] + args, cwd=cwd,
|
||||
capture_output=True, text=True, timeout=60)
|
||||
return (r.stdout + r.stderr).strip() or "(no output)"
|
||||
|
||||
@mcp.tool()
|
||||
def git_status(repo: str = "") -> str:
|
||||
"""Show git status of a repo (default: /workspace)."""
|
||||
return _git(["status", "--short"], repo)
|
||||
|
||||
@mcp.tool()
|
||||
def git_diff(repo: str = "", cached: bool = False) -> str:
|
||||
"""Show git diff (staged if cached=True)."""
|
||||
args = ["diff", "--stat", "--cached"] if cached else ["diff", "--stat"]
|
||||
return _git(args, repo) + "\n\n" + _git(
|
||||
["diff", "--cached"] if cached else ["diff"], repo)
|
||||
|
||||
@mcp.tool()
|
||||
def git_log(repo: str = "", n: int = 10) -> str:
|
||||
"""Show last n git commits."""
|
||||
return _git(["log", f"-{n}", "--oneline", "--decorate"], repo)
|
||||
|
||||
@mcp.tool()
|
||||
def git_commit(message: str, repo: str = "", add_all: bool = True) -> str:
|
||||
"""Stage all changes and create a commit."""
|
||||
if add_all:
|
||||
_git(["add", "-A"], repo)
|
||||
return _git(["commit", "-m", message], repo)
|
||||
|
||||
@mcp.tool()
|
||||
def git_checkout(branch: str, repo: str = "", create: bool = False) -> str:
|
||||
"""Checkout a branch, optionally creating it."""
|
||||
args = ["checkout", "-b", branch] if create else ["checkout", branch]
|
||||
return _git(args, repo)
|
||||
|
||||
# ── Gitea API ─────────────────────────────────────────────────────────────────
|
||||
def _gitea(method: str, path: str, body: dict = {}) -> dict:
|
||||
if not GITEA_TOKEN:
|
||||
return {"error": "GITEA_TOKEN not set in .env"}
|
||||
url = f"{GITEA_URL}/api/v1{path}"
|
||||
headers = {"Authorization": f"token {GITEA_TOKEN}",
|
||||
"Content-Type": "application/json"}
|
||||
r = httpx.request(method, url, json=body or None, headers=headers, timeout=30)
|
||||
try:
|
||||
return r.json()
|
||||
except Exception:
|
||||
return {"status": r.status_code, "text": r.text}
|
||||
|
||||
@mcp.tool()
|
||||
def gitea_list_repos() -> str:
|
||||
"""List your Gitea repos."""
|
||||
repos = _gitea("GET", "/repos/search?limit=50")
|
||||
if "error" in repos:
|
||||
return repos["error"]
|
||||
return "\n".join(f"{r['full_name']} — {r.get('description','')}"
|
||||
for r in repos.get("data", []))
|
||||
|
||||
@mcp.tool()
|
||||
def gitea_create_repo(name: str, private: bool = True, description: str = "") -> str:
|
||||
"""Create a new Gitea repository."""
|
||||
r = _gitea("POST", "/user/repos",
|
||||
{"name": name, "private": private, "description": description,
|
||||
"auto_init": True, "default_branch": "main"})
|
||||
return r.get("html_url") or str(r)
|
||||
|
||||
@mcp.tool()
|
||||
def gitea_create_issue(repo: str, title: str, body: str = "") -> str:
|
||||
"""Create an issue on a Gitea repo (format: owner/repo)."""
|
||||
r = _gitea("POST", f"/repos/{repo}/issues", {"title": title, "body": body})
|
||||
return r.get("html_url") or str(r)
|
||||
|
||||
# ── GitHub API ────────────────────────────────────────────────────────────────
|
||||
@mcp.tool()
|
||||
def github_api(method: str, endpoint: str, body: str = "") -> str:
|
||||
"""Call the GitHub REST API. endpoint e.g. /repos/owner/repo/issues"""
|
||||
if not GITHUB_TOKEN:
|
||||
return "GITHUB_TOKEN not set in .env"
|
||||
import json as _json
|
||||
headers = {"Authorization": f"Bearer {GITHUB_TOKEN}",
|
||||
"Accept": "application/vnd.github+json"}
|
||||
r = httpx.request(method.upper(), f"https://api.github.com{endpoint}",
|
||||
json=_json.loads(body) if body else None,
|
||||
headers=headers, timeout=30)
|
||||
try:
|
||||
return _json.dumps(r.json(), indent=2)
|
||||
except Exception:
|
||||
return r.text
|
||||
|
||||
# ── RAG ingest ────────────────────────────────────────────────────────────────
|
||||
@mcp.tool()
|
||||
def ingest_repo(url: str, name: str = "", branch: str = "main") -> str:
|
||||
"""Clone a git repo and index it in the RAG code collection."""
|
||||
r = httpx.post(f"{RAG_URL}/ingest/repo",
|
||||
json={"url": url, "name": name, "branch": branch}, timeout=300)
|
||||
return r.text
|
||||
|
||||
@mcp.tool()
|
||||
def rag_health() -> str:
|
||||
"""Check RAG server status and indexed document counts."""
|
||||
try:
|
||||
r = httpx.get(f"{RAG_URL}/health", timeout=10)
|
||||
return r.text
|
||||
except Exception as e:
|
||||
return f"RAG server unreachable: {e}"
|
||||
|
||||
if __name__ == "__main__":
|
||||
import uvicorn
|
||||
app = mcp.get_asgi_app()
|
||||
uvicorn.run(app, host="0.0.0.0", port=8002)
|
||||
@@ -0,0 +1,281 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
RAG Server — code-aware chunking, multi-collection, repo ingest, webhooks.
|
||||
Collections: papers (PDFs/text), code (source files, AST-split for Python)
|
||||
"""
|
||||
import ast, fnmatch, hashlib, json, logging, os, re, subprocess, threading, time
|
||||
from pathlib import Path
|
||||
from typing import Any, Optional
|
||||
|
||||
import chromadb
|
||||
import httpx
|
||||
from chromadb.utils.embedding_functions import OllamaEmbeddingFunction
|
||||
from fastapi import FastAPI, HTTPException, Request, BackgroundTasks
|
||||
from fastapi.middleware.cors import CORSMiddleware
|
||||
from fastapi.responses import StreamingResponse
|
||||
from pydantic import BaseModel
|
||||
|
||||
logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(message)s")
|
||||
log = logging.getLogger("rag")
|
||||
|
||||
OLLAMA_URL = os.getenv("OLLAMA_URL", "http://ollama:11434")
|
||||
CHROMA_URL = os.getenv("CHROMA_URL", "http://chromadb:8000")
|
||||
EMBED_MODEL = os.getenv("EMBED_MODEL", "nomic-embed-text")
|
||||
CHAT_MODEL = os.getenv("CHAT_MODEL", "qwen2.5:14b")
|
||||
PAPERS_DIR = Path(os.getenv("PAPERS_DIR", "/papers"))
|
||||
REPOS_DIR = Path(os.getenv("REPOS_DIR", "/repos"))
|
||||
TOP_K = int(os.getenv("TOP_K", "6"))
|
||||
|
||||
CODE_EXTS = {".py",".js",".ts",".tsx",".jsx",".go",".rs",".java",".c",".cpp",
|
||||
".h",".hpp",".cs",".rb",".sh",".yaml",".yml",".toml",".sql",".md"}
|
||||
SKIP_DIRS = {"node_modules",".git","__pycache__","dist","build",".venv",
|
||||
"venv","env",".next","vendor","target","bin","obj"}
|
||||
SKIP_FILES = {"package-lock.json","yarn.lock","pnpm-lock.yaml","Cargo.lock"}
|
||||
MAX_BYTES = 400_000
|
||||
|
||||
app = FastAPI(title="RAG Server")
|
||||
app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])
|
||||
|
||||
# ── ChromaDB ──────────────────────────────────────────────────────────────────
|
||||
def _embed_fn():
|
||||
return OllamaEmbeddingFunction(
|
||||
url=f"{OLLAMA_URL}/api/embeddings", model_name=EMBED_MODEL)
|
||||
|
||||
def _chroma():
|
||||
host, port = CHROMA_URL.replace("http://","").split(":")
|
||||
return chromadb.HttpClient(host=host, port=int(port))
|
||||
|
||||
def get_col(name: str):
|
||||
return _chroma().get_or_create_collection(name, embedding_function=_embed_fn())
|
||||
|
||||
# ── chunkers ─────────────────────────────────────────────────────────────────
|
||||
def _doc_id(text: str, key: str) -> str:
|
||||
return hashlib.md5(f"{key}|{text[:200]}".encode()).hexdigest()
|
||||
|
||||
def _sliding(text: str, size=1000, overlap=150) -> list[str]:
|
||||
chunks, i = [], 0
|
||||
while i < len(text):
|
||||
chunks.append(text[i:i+size])
|
||||
i += size - overlap
|
||||
return [c for c in chunks if c.strip()]
|
||||
|
||||
def _chunk_python(src: str) -> list[tuple[str,str]]:
|
||||
try:
|
||||
tree = ast.parse(src)
|
||||
except SyntaxError:
|
||||
return []
|
||||
lines = src.splitlines()
|
||||
out = []
|
||||
for node in ast.iter_child_nodes(tree):
|
||||
if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):
|
||||
chunk = "\n".join(lines[node.lineno-1:node.end_lineno])
|
||||
out.append((node.name, chunk[:4000]))
|
||||
return out
|
||||
|
||||
def _chunk_file(path: Path, src: str) -> list[tuple[str,str]]:
|
||||
if path.suffix == ".py":
|
||||
pairs = _chunk_python(src)
|
||||
if pairs:
|
||||
return pairs
|
||||
# function/class boundary split for JS/TS/Go/Rust etc.
|
||||
pat = re.compile(
|
||||
r'(?:^|\n)(?=(?:export\s+)?(?:async\s+)?(?:function|class|const\s+\w+\s*=\s*(?:async\s+)?\()'
|
||||
r'|^func |^type |^impl |^pub fn |^fn )',
|
||||
re.MULTILINE)
|
||||
parts = [p.strip() for p in pat.split(src) if p.strip()]
|
||||
if len(parts) > 1:
|
||||
return [(f"s{i}", p[:4000]) for i, p in enumerate(parts)]
|
||||
return [(f"c{i}", c) for i, c in enumerate(_sliding(src, 1200, 200))]
|
||||
|
||||
# ── ingest helpers ────────────────────────────────────────────────────────────
|
||||
def ingest_file(col, fpath: Path, repo: str = ""):
|
||||
if fpath.stat().st_size > MAX_BYTES or fpath.name in SKIP_FILES:
|
||||
return
|
||||
if any(fnmatch.fnmatch(fpath.name, p) for p in ("*.min.js","*.min.css","*.map")):
|
||||
return
|
||||
try:
|
||||
src = fpath.read_text(encoding="utf-8", errors="ignore")
|
||||
except Exception:
|
||||
return
|
||||
if not src.strip():
|
||||
return
|
||||
rel = str(fpath)
|
||||
pairs = _chunk_file(fpath, src) if fpath.suffix in CODE_EXTS else \
|
||||
[(f"c{i}", c) for i, c in enumerate(_sliding(src))]
|
||||
ids, docs, metas = [], [], []
|
||||
for label, chunk in pairs:
|
||||
if not chunk.strip():
|
||||
continue
|
||||
ids.append(_doc_id(chunk, rel+label))
|
||||
docs.append(chunk)
|
||||
metas.append({"source": rel, "label": label, "repo": repo,
|
||||
"lang": fpath.suffix.lstrip(".")})
|
||||
if ids:
|
||||
col.upsert(ids=ids, documents=docs, metadatas=metas)
|
||||
|
||||
def ingest_dir(col, directory: Path, repo: str = "") -> int:
|
||||
count = 0
|
||||
for f in directory.rglob("*"):
|
||||
if not f.is_file():
|
||||
continue
|
||||
if any(p in f.parts for p in SKIP_DIRS):
|
||||
continue
|
||||
ingest_file(col, f, repo)
|
||||
count += 1
|
||||
log.info("Indexed %d files from %s", count, directory)
|
||||
return count
|
||||
|
||||
def ingest_pdfs(col) -> int:
|
||||
try:
|
||||
import pypdf
|
||||
except ImportError:
|
||||
log.warning("pypdf not installed — skipping PDFs")
|
||||
return 0
|
||||
n = 0
|
||||
for pdf in PAPERS_DIR.glob("*.pdf"):
|
||||
try:
|
||||
text = "\n".join(p.extract_text() or ""
|
||||
for p in pypdf.PdfReader(str(pdf)).pages)
|
||||
for i, chunk in enumerate(_sliding(text)):
|
||||
col.upsert(ids=[_doc_id(chunk, str(pdf)+str(i))],
|
||||
documents=[chunk],
|
||||
metadatas=[{"source": str(pdf), "label": f"p{i}",
|
||||
"repo": "", "lang": "pdf"}])
|
||||
n += 1
|
||||
except Exception as e:
|
||||
log.warning("PDF %s: %s", pdf.name, e)
|
||||
return n
|
||||
|
||||
# ── startup ───────────────────────────────────────────────────────────────────
|
||||
def _startup_index():
|
||||
# wait for embed model
|
||||
for _ in range(40):
|
||||
try:
|
||||
r = httpx.get(f"{OLLAMA_URL}/api/tags", timeout=5)
|
||||
if any(EMBED_MODEL in m["name"] for m in r.json().get("models", [])):
|
||||
break
|
||||
except Exception:
|
||||
pass
|
||||
log.info("Waiting for embed model %s…", EMBED_MODEL)
|
||||
time.sleep(5)
|
||||
|
||||
code_col = get_col("code")
|
||||
papers_col = get_col("papers")
|
||||
for d in REPOS_DIR.iterdir():
|
||||
if d.is_dir():
|
||||
ingest_dir(code_col, d, d.name)
|
||||
ingest_pdfs(papers_col)
|
||||
for f in PAPERS_DIR.glob("*.txt"):
|
||||
ingest_file(papers_col, f)
|
||||
log.info("Startup index complete")
|
||||
|
||||
@app.on_event("startup")
|
||||
async def on_startup():
|
||||
threading.Thread(target=_startup_index, daemon=True).start()
|
||||
|
||||
# ── endpoints ─────────────────────────────────────────────────────────────────
|
||||
@app.get("/health")
|
||||
async def health():
|
||||
try:
|
||||
cc = _chroma()
|
||||
return {"status": "ok",
|
||||
"code": cc.get_collection("code", embedding_function=_embed_fn()).count(),
|
||||
"papers": cc.get_collection("papers", embedding_function=_embed_fn()).count(),
|
||||
"embed": EMBED_MODEL, "chat": CHAT_MODEL}
|
||||
except Exception as e:
|
||||
return {"status": "error", "detail": str(e)}
|
||||
|
||||
class RepoRequest(BaseModel):
|
||||
url: str
|
||||
name: str = ""
|
||||
branch: str = "main"
|
||||
|
||||
@app.post("/ingest/repo")
|
||||
async def ingest_repo(req: RepoRequest):
|
||||
name = req.name or req.url.rstrip("/").split("/")[-1].removesuffix(".git")
|
||||
dest = REPOS_DIR / name
|
||||
try:
|
||||
if dest.exists():
|
||||
subprocess.run(["git","pull"], cwd=dest, check=True, timeout=120)
|
||||
else:
|
||||
subprocess.run(["git","clone","--depth=1","-b",req.branch,
|
||||
req.url, str(dest)], check=True, timeout=300)
|
||||
except subprocess.CalledProcessError as e:
|
||||
raise HTTPException(400, str(e))
|
||||
n = ingest_dir(get_col("code"), dest, name)
|
||||
return {"status": "ok", "repo": name, "files": n}
|
||||
|
||||
@app.post("/ingest/papers")
|
||||
async def trigger_papers(bg: BackgroundTasks):
|
||||
bg.add_task(ingest_pdfs, get_col("papers"))
|
||||
return {"status": "queued"}
|
||||
|
||||
async def _webhook(payload: dict):
|
||||
repo = payload.get("repository") or {}
|
||||
url = repo.get("clone_url") or repo.get("html_url","")
|
||||
name = repo.get("name","unknown")
|
||||
if not url:
|
||||
return {"status": "ignored"}
|
||||
dest = REPOS_DIR / name
|
||||
if dest.exists():
|
||||
subprocess.run(["git","pull"], cwd=dest, timeout=120)
|
||||
else:
|
||||
subprocess.run(["git","clone","--depth=1",url,str(dest)], timeout=300)
|
||||
n = ingest_dir(get_col("code"), dest, name)
|
||||
return {"status": "ok", "repo": name, "files": n}
|
||||
|
||||
@app.post("/webhook/gitea")
|
||||
async def webhook_gitea(r: Request): return await _webhook(await r.json())
|
||||
|
||||
@app.post("/webhook/github")
|
||||
async def webhook_github(r: Request): return await _webhook(await r.json())
|
||||
|
||||
# ── RAG chat ──────────────────────────────────────────────────────────────────
|
||||
class ChatRequest(BaseModel):
|
||||
model: str = CHAT_MODEL
|
||||
messages: list[dict[str,Any]]
|
||||
stream: bool = False
|
||||
collections: list[str] = ["code","papers"]
|
||||
|
||||
def _context(query: str, cols: list[str]) -> str:
|
||||
parts = []
|
||||
for cname in cols:
|
||||
try:
|
||||
col = get_col(cname)
|
||||
if col.count() == 0:
|
||||
continue
|
||||
res = col.query(query_texts=[query], n_results=min(TOP_K, col.count()))
|
||||
for doc, meta in zip(res["documents"][0], res["metadatas"][0]):
|
||||
parts.append(f"### {meta.get('source','')}:{meta.get('label','')}\n"
|
||||
f"```{meta.get('lang','')}\n{doc}\n```")
|
||||
except Exception as e:
|
||||
log.warning("col %s: %s", cname, e)
|
||||
return "\n\n".join(parts)
|
||||
|
||||
@app.post("/v1/chat/completions")
|
||||
async def chat(req: ChatRequest):
|
||||
query = next((m["content"] for m in reversed(req.messages)
|
||||
if m.get("role")=="user"), "")
|
||||
context = _context(query, req.collections)
|
||||
msgs = [{"role":"system","content":
|
||||
"You are a helpful coding assistant. Use the retrieved context below.\n\n"
|
||||
f"## Context\n{context}"}] + req.messages
|
||||
|
||||
payload = {"model": req.model, "messages": msgs, "stream": req.stream}
|
||||
|
||||
if req.stream:
|
||||
async def gen():
|
||||
async with httpx.AsyncClient(timeout=300) as client:
|
||||
async with client.stream("POST",
|
||||
f"{OLLAMA_URL}/v1/chat/completions", json=payload) as r:
|
||||
async for chunk in r.aiter_bytes():
|
||||
yield chunk
|
||||
return StreamingResponse(gen(), media_type="text/event-stream")
|
||||
|
||||
async with httpx.AsyncClient(timeout=300) as client:
|
||||
r = await client.post(f"{OLLAMA_URL}/v1/chat/completions", json=payload)
|
||||
return r.json()
|
||||
|
||||
if __name__ == "__main__":
|
||||
import uvicorn
|
||||
uvicorn.run("server:app", host="0.0.0.0", port=8001, reload=False)
|
||||
Reference in New Issue
Block a user