Agent is not using my documents

[nachogarciam]

Describe the bug
When selecting "my documents" on UI, it does not use my documents, it uses the search engine.

To Reproduce
Steps to reproduce the behavior:

1. clone repo
2. config keys (im using azure openai model)
3. upload a txt file to "./my-docs" folder
4. run app with the specified command.
5. In sources selection section select "my documents"
6. ask

Expected behavior
I expect that the model read my documents and not search on internet.

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: windows
• Browser MS Edge

Additional context
Add any other context about the problem here.
Im using the same model for "FAST", "SMART" and "STRATEGIC_LLM"

FAST_LLM=azure_openai:gpt-4o-mini
SMART_LLM=azure_openai:gpt-4o-mini
STRATEGIC_LLM=azure_openai:gpt-4o-mini

EMBEDDING="azure_openai:text-embedding-ada-002"

RESULT:

[adrianhensler]

Just because you did not mention it, and it is showing in the image you shared - may I ask if you set the doc_path?

I believe this is required on Windows as well as linux (I run it on a chromebook most of the time, in Ubuntu)
https://docs.gptr.dev/docs/gpt-researcher/context/tailored-research#research-on-local-documents-

[nachogarciam]

Hi Adrian, yes I did set the path

[knoel99]

I have the same behavior, when the app go search on the web, even if I can see my documents in my path in Preferences > Report Source > My Documents

[fan-wen]

I have the same behavior too - this might be a bug...

[amirsa66]

@nachogarciam
This bug doesn't fix yet?

[lmontin]

I also experience this problem:

I have uploaded a file and set Report Source as "My Documents" but it immediately starts a Web Search

gpt-researcher-1 | Files in ./my-docs: ['abc.pdf']
gpt-researcher-1 | INFO: 172.19.0.1:48572 - "GET /files/ HTTP/1.1" 200 OK
gpt-researcher-1 | INFO: ('172.19.0.1', 48586) - "WebSocket /ws" [accepted]
gpt-researcher-1 | INFO: connection open
gpt-researcher-1 | INFO: connection closed
gpt-researcher-1 | INFO: ('172.19.0.1', 57300) - "WebSocket /ws" [accepted]
gpt-researcher-1 | INFO: connection open
gpt-researcher-1 | 2025-01-07 08:43:48,280 - research - INFO - Starting research for query: summarize document

gpt-researcher-1 | INFO: [08:43:48] 🔍 Starting the research task for 'summarize document'...

gpt-researcher-1 | INFO: [08:43:48] 📝 Document Analysis Agent

gpt-researcher-1 | 2025-01-07 08:43:48,281 - research - INFO - Using web search

gpt-researcher-1 | 2025-01-07 08:43:48,281 - research - INFO - Starting web search for query: summarize document

[lmontin]

i did some digging and it seems the form report_source value is not being passed or is being overrided so that self.researcher.report_source always equals "web" and does not trigger the "local" option in gpt-researcher > gpt_researcher > skills > researcher.py > ResearchConductor > conduct_research
I tried to find why but couldn't figure out the problem .

[alesaccoia]

Fixed with PR #1070

[realannieqiu]

Hi, I have the same issue even after the fix. The documents are loaded but the web search starts immediately afterwards, and no reference to the local documents is made in the report. Any idea what the error is here? Thank you.