Issue with DuckDuckGo in knowledge tool

[alvaromongon]

Hello, not sure this is a miss-configuration on my side.

I don´t have a perplexity api key and therefore the knowledge tool only uses duckduckgo search. But it is not working. checking the output I am always getting a RateLimit error.

I could not see any api key for that neither ratelimits for using duckduck go search online.

Any idea what to do here?

I was also thinking, if there is a web page content tool running queries against search engines. Could we build something similar to that for the knowledge tool? I guess the main issue is parsing the result?

[warclone]

I commented out the part of the ddg query that kept failing. Couldn't tell if a0 ever successfully used it.

[FarVision2]

I dropped in a SEARXNG docker instance and told it to use that and forget all about DDG. Had it run a handful of search tests to 'learn' that it worked. (because it used to try to use DDG to search for SEARXNG)

I used to build my own images (https://github.com/searxng/searxng)

Then took the easy way out and ran the compose (https://github.com/searxng/searxng-docker)

Utility Message:
6 entries memorized.

Memories:
Local SEARXNG instances: http://127.0.0.1:8380/, http://host.docker.internal:8380/
SEARXNG instance selection: Use http://127.0.0.1:8380/ if Docker sandbox is not running; otherwise, use http://host.docker.internal:8380/

Replaced:
id: f6ad5127-932b-4ff0-9c7c-d46f9e7f88c4
timestamp: 2024-10-20 12:31:16
area: fragments
Content: SEARXNG search URLs: http://127.0.0.1:8380/, http://host.docker.internal:8380/

Agent 2: Using tool 'knowledge_tool'
Question    
What is the correct pip package name for the DeepSORT object tracking library?

# Memory
id: e81c844c-f6f4-44ca-8722-965ffa6e917b
timestamp: 2024-10-20 12:43:49
area: fragments
Content: Local SEARXNG instances: http://127.0.0.1:8380/, http://host.docker.internal:8380/

id: a22baa9d-d3c7-4562-8f53-1da4528ad0ae
timestamp: 2024-10-20 12:43:50
area: fragments
Content: SEARXNG instance selection: Use http://127.0.0.1:8380/ if Docker sandbox is not running; otherwise, use http://host.docker.internal:8380/

No API key provided for Perplexity. Skipping Perplexity search.

Agent 2: Generating
Thoughts    
The knowledge_tool search for the DeepSORT package name failed due to rate limiting.
I will use my local SEARXNG instance to search for the correct package name.
I will then install the correct package using pip and rerun the code.
Reflection  
Using a local search engine avoids rate limits and ensures faster responses.
The failure of the initial search highlights the importance of having alternative search methods.
The next step is to thoroughly test the pipeline with a sample video to evaluate its performance and identify any further issues.
Tool Name   
code_execution_tool
Runtime 
terminal
Code    
pip install deepsort

Everything else performs as expected.

[cal88]

The DDGS package version used in A0 somehow has a very low rate limit. Upgrading it to the newest version seems to solve it.