Remove test_channel_by_id_or_name.py
- Clean up temporary test file that was used for debugging - Keep only the main slack_rag.py application for production use
This commit is contained in:
@@ -1,183 +0,0 @@
|
||||
import argparse
|
||||
import asyncio
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
sys.path.insert(0, str(Path(__file__).parent / "apps"))
|
||||
from slack_data.slack_mcp_reader import SlackMCPReader
|
||||
|
||||
|
||||
async def fetch(
|
||||
channel_id: str | None,
|
||||
channel_name: str | None,
|
||||
workspace_name: str | None,
|
||||
query: str,
|
||||
search: str | None,
|
||||
):
|
||||
reader = SlackMCPReader(
|
||||
mcp_server_command="slack-mcp-server",
|
||||
workspace_name=workspace_name,
|
||||
concatenate_conversations=True,
|
||||
max_messages_per_conversation=10000000000,
|
||||
max_retries=5,
|
||||
retry_delay=2.0,
|
||||
)
|
||||
async with reader:
|
||||
print("Connected to Slack MCP server!")
|
||||
if channel_name and not channel_id:
|
||||
lst = await reader.send_mcp_request(
|
||||
{
|
||||
"jsonrpc": "2.0",
|
||||
"id": 1,
|
||||
"method": "tools/call",
|
||||
"params": {"name": "channels_list", "arguments": {}},
|
||||
}
|
||||
)
|
||||
text = lst.get("result", {}).get("content", [{"text": ""}])[0]["text"]
|
||||
for line in text.splitlines()[1:]:
|
||||
if not line.strip():
|
||||
continue
|
||||
parts = [p.strip() for p in line.split(",")]
|
||||
if len(parts) < 2:
|
||||
continue
|
||||
cid, name = parts[0], parts[1].lstrip("#")
|
||||
if name.lower() == channel_name.lower():
|
||||
channel_id = cid
|
||||
print(f"Resolved channel name #{channel_name} -> {channel_id}")
|
||||
break
|
||||
if not channel_id:
|
||||
print(f"No channel named '{channel_name}' found.")
|
||||
return
|
||||
if not channel_id:
|
||||
print("Provide --channel-id or --channel-name.")
|
||||
return
|
||||
|
||||
# If search is provided, try to use a search tool first
|
||||
resp = None
|
||||
if search:
|
||||
try:
|
||||
tools = await reader.list_available_tools()
|
||||
search_tool = None
|
||||
for t in tools:
|
||||
name = t.get("name", "").lower()
|
||||
if "search" in name and "message" in name:
|
||||
search_tool = t["name"]
|
||||
break
|
||||
if search_tool:
|
||||
print(f"Searching with tool '{search_tool}' for: {search}")
|
||||
resp = await reader.send_mcp_request(
|
||||
{
|
||||
"jsonrpc": "2.0",
|
||||
"id": 2,
|
||||
"method": "tools/call",
|
||||
"params": {
|
||||
"name": search_tool,
|
||||
"arguments": {
|
||||
"query": search,
|
||||
"channel_id": channel_id,
|
||||
"limit": 200,
|
||||
},
|
||||
},
|
||||
}
|
||||
)
|
||||
else:
|
||||
print("Search tool not available, falling back to full history.")
|
||||
except Exception as e:
|
||||
print(f"Search failed ({e}), falling back to full history.")
|
||||
|
||||
if resp is None:
|
||||
print(f"Fetching messages from {channel_id} ...")
|
||||
resp = await reader.send_mcp_request(
|
||||
{
|
||||
"jsonrpc": "2.0",
|
||||
"id": 2,
|
||||
"method": "tools/call",
|
||||
"params": {
|
||||
"name": "conversations_history",
|
||||
"arguments": {"channel_id": channel_id, "limit": 10000000000},
|
||||
},
|
||||
}
|
||||
)
|
||||
if "error" in resp:
|
||||
msg = resp["error"].get("message", "Unknown")
|
||||
print("Error:", msg)
|
||||
if msg in ("not_in_channel", "channel_not_found"):
|
||||
print("Tip: invite the bot to the channel: /invite @YourBotName")
|
||||
return
|
||||
result = resp.get("result", {})
|
||||
content = result.get("content")
|
||||
text_blob = None
|
||||
if isinstance(content, list) and content and "text" in content[0]:
|
||||
text_blob = content[0]["text"]
|
||||
print(text_blob[:4000])
|
||||
else:
|
||||
print(result)
|
||||
|
||||
# Simple RAG-style answer with LEANN-focused boosting
|
||||
if query and text_blob:
|
||||
print("\n" + "=" * 60)
|
||||
print("RAG ANSWER")
|
||||
print("=" * 60)
|
||||
q_terms = [t.strip().lower() for t in query.split() if t.strip()]
|
||||
lines = [
|
||||
line
|
||||
for line in (text_blob.splitlines() if text_blob else [])
|
||||
if line and not line.startswith("MsgID,")
|
||||
]
|
||||
# Score lines by count of query terms present
|
||||
scored = []
|
||||
boost_terms = {
|
||||
"leann": 5,
|
||||
"yichuan-w/leann": 4,
|
||||
"github.com/yichuan-w/leann": 4,
|
||||
"x.com/yichuanm": 3,
|
||||
"leann vector": 3,
|
||||
}
|
||||
for ln in lines:
|
||||
ll = ln.lower()
|
||||
score = sum(1 for t in q_terms if t in ll)
|
||||
for k, b in boost_terms.items():
|
||||
if k in ll:
|
||||
score += b
|
||||
if score > 0:
|
||||
scored.append((score, ln))
|
||||
scored.sort(key=lambda x: x[0], reverse=True)
|
||||
top = [ln for _, ln in scored[:5]]
|
||||
if top:
|
||||
print(f"Query: {query}")
|
||||
print("Relevant messages:")
|
||||
for i, ln in enumerate(top, 1):
|
||||
print(f" {i}. {ln}")
|
||||
leann_hits = [ln for ln in lines if "leann" in ln.lower()][:5]
|
||||
if leann_hits:
|
||||
print("\nLEANN-focused highlights:")
|
||||
for i, ln in enumerate(leann_hits, 1):
|
||||
print(f" {i}. {ln}")
|
||||
else:
|
||||
print(f"Query: {query}")
|
||||
print("No directly matching messages found; showing recent context:")
|
||||
for i, ln in enumerate(lines[:5], 1):
|
||||
print(f" {i}. {ln}")
|
||||
|
||||
|
||||
def main():
|
||||
ap = argparse.ArgumentParser()
|
||||
ap.add_argument("--channel-id", default=None)
|
||||
ap.add_argument("--channel-name", default=None, help="e.g., random")
|
||||
ap.add_argument("--workspace-name", default="Sky Lab Computing")
|
||||
ap.add_argument("--query", default="What is LEANN about?", help="Simple RAG-style query")
|
||||
ap.add_argument("--search", default=None, help="Server-side message search query")
|
||||
args = ap.parse_args()
|
||||
asyncio.run(
|
||||
fetch(
|
||||
args.channel_id,
|
||||
args.channel_name,
|
||||
args.workspace_name,
|
||||
args.query,
|
||||
args.search,
|
||||
)
|
||||
)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
Reference in New Issue
Block a user