218 lines
7.1 KiB
Python
218 lines
7.1 KiB
Python
import json
|
|
import logging
|
|
import os
|
|
from datetime import datetime, timezone
|
|
from pathlib import Path
|
|
|
|
import httpx
|
|
from fastapi import FastAPI, HTTPException
|
|
from pydantic import BaseModel
|
|
|
|
logging.basicConfig(level=logging.INFO, format="%(asctime)s [sam] %(message)s")
|
|
logger = logging.getLogger("sam")
|
|
|
|
app = FastAPI(title="sam-research", version="0.1.0")
|
|
|
|
OLLAMA_URL = os.getenv("OLLAMA_URL", "http://172.27.6.139:11434")
|
|
OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "gemma4")
|
|
SEARXNG_URL = os.getenv("SEARXNG_URL", "http://searxng:8080")
|
|
SITES_DIR = Path(os.getenv("SITES_DIR", "/opt/sites"))
|
|
AGENT_OS_DIR = Path(os.getenv("AGENT_OS_DIR", "/opt/agent-os"))
|
|
|
|
HISTORY_FILE = SITES_DIR / "sam" / "history.json"
|
|
MAX_HISTORY = 10
|
|
|
|
|
|
def load_history() -> list[dict]:
|
|
if HISTORY_FILE.exists():
|
|
try:
|
|
return json.loads(HISTORY_FILE.read_text())
|
|
except Exception:
|
|
pass
|
|
return []
|
|
|
|
|
|
def save_history(entry: dict):
|
|
history = load_history()
|
|
history.insert(0, entry)
|
|
HISTORY_FILE.parent.mkdir(parents=True, exist_ok=True)
|
|
HISTORY_FILE.write_text(json.dumps(history[:MAX_HISTORY], indent=2))
|
|
|
|
|
|
async def search_web(query: str, max_results: int) -> list[dict]:
|
|
params = {"q": query, "format": "json", "categories": "general", "language": "en"}
|
|
async with httpx.AsyncClient(timeout=30.0) as client:
|
|
r = await client.get(f"{SEARXNG_URL}/search", params=params)
|
|
r.raise_for_status()
|
|
data = r.json()
|
|
return [
|
|
{"title": item.get("title", ""), "url": item.get("url", ""), "snippet": item.get("content", "")}
|
|
for item in data.get("results", [])[:max_results]
|
|
]
|
|
|
|
|
|
async def ollama_synthesize(prompt: str, model: str) -> str:
|
|
payload = {
|
|
"model": model,
|
|
"messages": [{"role": "user", "content": prompt}],
|
|
"stream": False,
|
|
}
|
|
async with httpx.AsyncClient(timeout=120.0) as client:
|
|
r = await client.post(f"{OLLAMA_URL}/api/chat", json=payload)
|
|
r.raise_for_status()
|
|
return r.json()["message"]["content"]
|
|
|
|
|
|
def build_prompt(query: str, results: list[dict], instructions: str | None) -> str:
|
|
lines = [f"You are a research assistant. The user wants information on: {query}\n"]
|
|
if instructions:
|
|
lines.append(f"Instructions: {instructions}\n")
|
|
lines.append("Web search results:\n")
|
|
for i, r in enumerate(results, 1):
|
|
lines.append(f"{i}. **{r['title']}**")
|
|
lines.append(f" URL: {r['url']}")
|
|
if r["snippet"]:
|
|
lines.append(f" {r['snippet']}")
|
|
lines.append("")
|
|
lines.append(
|
|
"Based on these results, provide a thorough answer. "
|
|
"Follow any format instructions given above. "
|
|
"Cite sources by referencing their URLs inline."
|
|
)
|
|
return "\n".join(lines)
|
|
|
|
|
|
def render_html(history: list[dict]) -> str:
|
|
now_iso = datetime.now(timezone.utc).isoformat()
|
|
|
|
rows = ""
|
|
for entry in history:
|
|
ts = entry.get("timestamp", "")
|
|
ts_span = f'<span data-utc="{ts}"></span>' if ts else "—"
|
|
query = entry.get("query", "")
|
|
model = entry.get("model", "")
|
|
result_len = entry.get("summary_chars", 0)
|
|
rows += f"""
|
|
<tr>
|
|
<td>{ts_span}</td>
|
|
<td class="query">{query}</td>
|
|
<td class="dim">{model}</td>
|
|
<td class="dim">{result_len} chars</td>
|
|
</tr>"""
|
|
|
|
empty = "<tr><td colspan='4' class='dim'>No research runs yet.</td></tr>" if not rows else rows
|
|
|
|
return f"""<!DOCTYPE html>
|
|
<html lang="en">
|
|
<head>
|
|
<meta charset="UTF-8">
|
|
<meta name="viewport" content="width=device-width, initial-scale=1">
|
|
<title>Sam — Research</title>
|
|
<style>
|
|
* {{ box-sizing: border-box; margin: 0; padding: 0; }}
|
|
body {{ font-family: monospace; background: #0d1117; color: #c9d1d9; padding: 2rem; }}
|
|
h1 {{ color: #58a6ff; margin-bottom: 0.25rem; }}
|
|
.desc {{ color: #8b949e; font-size: 0.9rem; margin: 0.4rem 0 0.25rem; }}
|
|
.meta {{ color: #8b949e; font-size: 0.85rem; margin-bottom: 2rem; }}
|
|
.meta a {{ color: #8b949e; }}
|
|
h2 {{ color: #8b949e; font-size: 0.85rem; text-transform: uppercase; letter-spacing: 0.05em; margin: 2rem 0 0.75rem; }}
|
|
table {{ width: 100%; border-collapse: collapse; }}
|
|
td {{ padding: 0.6rem 1rem; border-bottom: 1px solid #21262d; font-size: 0.9rem; }}
|
|
.query {{ color: #c9d1d9; }}
|
|
.dim {{ color: #8b949e; }}
|
|
</style>
|
|
</head>
|
|
<body>
|
|
<h1>Sam — Research Agent</h1>
|
|
<p class="desc">Web search via SearXNG + synthesis via Ollama. POST /research to use.</p>
|
|
<p class="meta">Updated <span data-utc="{now_iso}"></span> · <a href="/">← home</a></p>
|
|
<h2>Recent Queries</h2>
|
|
<table><tbody>{empty}</tbody></table>
|
|
<script>
|
|
document.querySelectorAll('[data-utc]').forEach(el => {{
|
|
el.textContent = new Date(el.dataset.utc).toLocaleString(undefined, {{year:'numeric',month:'2-digit',day:'2-digit',hour:'2-digit',minute:'2-digit'}});
|
|
}});
|
|
</script>
|
|
</body>
|
|
</html>"""
|
|
|
|
|
|
class ResearchRequest(BaseModel):
|
|
query: str
|
|
max_results: int = 5
|
|
instructions: str | None = None
|
|
model: str | None = None
|
|
|
|
|
|
@app.get("/health")
|
|
def health():
|
|
return {
|
|
"status": "ok",
|
|
"agent": "sam-research",
|
|
"searxng_url": SEARXNG_URL,
|
|
"ollama_url": OLLAMA_URL,
|
|
"model": OLLAMA_MODEL,
|
|
}
|
|
|
|
|
|
@app.post("/research")
|
|
async def research(req: ResearchRequest):
|
|
model = req.model or OLLAMA_MODEL
|
|
timestamp = datetime.now(timezone.utc).isoformat()
|
|
|
|
logger.info(f"research: query='{req.query}' max_results={req.max_results} model={model}")
|
|
|
|
try:
|
|
results = await search_web(req.query, req.max_results)
|
|
except httpx.HTTPError as e:
|
|
raise HTTPException(status_code=502, detail=f"SearXNG error: {e}")
|
|
|
|
logger.info(f"got {len(results)} results from SearXNG")
|
|
|
|
prompt = build_prompt(req.query, results, req.instructions)
|
|
|
|
try:
|
|
summary = await ollama_synthesize(prompt, model)
|
|
except httpx.HTTPError as e:
|
|
raise HTTPException(status_code=502, detail=f"Ollama error: {e}")
|
|
|
|
logger.info(f"synthesis complete: {len(summary)} chars")
|
|
|
|
out_dir = SITES_DIR / "sam"
|
|
out_dir.mkdir(parents=True, exist_ok=True)
|
|
|
|
(out_dir / "last-output.md").write_text(
|
|
f"# Sam Research — {req.query}\n\n"
|
|
f"_Query: {req.query}_ \n"
|
|
f"_Model: {model}_ \n"
|
|
f"_Results: {len(results)}_ \n"
|
|
f"_Timestamp: {timestamp}_\n\n"
|
|
f"---\n\n{summary}"
|
|
)
|
|
|
|
history_entry = {
|
|
"timestamp": timestamp,
|
|
"query": req.query,
|
|
"model": model,
|
|
"summary_chars": len(summary),
|
|
}
|
|
save_history(history_entry)
|
|
(out_dir / "index.html").write_text(render_html(load_history()))
|
|
|
|
log_dir = AGENT_OS_DIR / "logs" / "sam-research"
|
|
log_dir.mkdir(parents=True, exist_ok=True)
|
|
(log_dir / "last-run.json").write_text(json.dumps({
|
|
"agent": "sam-research",
|
|
"timestamp": timestamp,
|
|
"status": "success",
|
|
"result": f"query='{req.query}' results={len(results)} summary={len(summary)} chars",
|
|
}, indent=2))
|
|
|
|
return {
|
|
"query": req.query,
|
|
"results": results,
|
|
"summary": summary,
|
|
"model": model,
|
|
"timestamp": timestamp,
|
|
}
|