Friends of the Crustacean ๐Ÿฆž๐ŸคFotC๏ฟฝ
Friends of the Crustacean ๐Ÿฆž๐Ÿคโ€ข3d agoโ€ข
54 replies
lirantal

tool calls dont return results, maybe its a session issue?

I get this issue often with tool calls like when reading files or now the brave search api that dont show back any results. I am not sure if this is maybe related to bad session setup.

Channel is telegram, here's an example transcript:

> me:
it doesn't seem like you were able to find the actual movies that air tonight in Cinema City in Brooklyn theater. Can you try a refined search for that?

> clawd:
Okay, I understand. The previous search gave general information but not the specific movie titles. I'll try a more refined search query to get closer to the actual movie listings.

[Tool Call: web_search (ID: web_search_371827391921_12)]
Arguments: {
"count": 5,
"query": "movies playing tonight after 9pm Cinema City Brooklyn"
}

> me:
did it not work ?

> clawd:
I'm waiting for the results of the refined search to come back. I'll let you know what I find as soon as it's available.


the tool result never returns

if I view clawdbot sessions I get these:

Session store: /home/node/.clawdbot/agents/main/sessions/sessions.json
Sessions listed: 2
Kind   Key                        Age       Model          Tokens (ctx %)       Flags
direct agent:main:main            4m ago    gemini-2.5-flash-lite 8.5k/1049k (1%)      system id:1111---
direct main                       24h ago   gemini-3-pro-preview 206k/1000k (21%)     id:222-----


I indeed moved to gemini-2.5 and it seems that all my convo with telegram is going through that "agent:main:main" session. shouldn't it be the main one?

I'm not sure if the clawdbot git repo is confusing the actual workspace.
- cloned git repo to ~/clawdbot-git and running pnpm clawdbot from there for one off commands but installed the gateway through user daemon
- my clawdbot dir for workspace should be ~/clawd from what I can tell
Was this page helpful?