Skip to content

Containerised Search Agent and UI#3

Open
VJonah wants to merge 10 commits into
mainfrom
agent-ui
Open

Containerised Search Agent and UI#3
VJonah wants to merge 10 commits into
mainfrom
agent-ui

Conversation

@VJonah
Copy link
Copy Markdown

@VJonah VJonah commented Dec 15, 2025

The search agent and its basic web UI are containerised for easy sharing and user testing.

jonah added 10 commits December 2, 2025 18:05
previously, the project relied on a python 3.11 minimum version. this conflicted with the requirements of destiny_sdk.

now, the project's minimum python version has been bumped to 3.12
previously, we created a paper helper for OpenAlex inspired by the existing one for OpenReview

now, we have a basic paper helper that searches the DESTINY repo to answer a user's question
previously, we set the paper_directory Settings field to "./literature" within the .json config

now, that no longer is the case
previously, we only had a DESTINY paper helper that could download the first page of references that resulted from an LLM generated search query

now, we have a dedicated DESTINY paper search tool an agent can use to source information for its other steps
previously, to use the new DESTINY search tool, we had to explicitly list the agent's available tools

now, we have a config that unifies our azure and agent tool configs in a single json that can more easily instantiate Settings objects
previously, we used settings.paper_directory to decide where pdfs of destiny references should be downloaded

now, since settings.paper_directory is deprecated, we use the more appropriate settings.agent.index.paper_directory instead
previously, we only detailed how to use the DESTINY paper helper

now, we briefly describe how to use the agent with the new DESTINY search tool
previously, running the paperqa agent could only be done via the terminal or a notebook

now, we have a basic web UI built with Chainlit that allows for a QA session to run in a temporary dictionary (no persistence yet)
previously, our basic UI only showed the agent's final response

now, with improved tool responses and bug fixes as well as callbacks to display intermediary tool calls as Chainlit Steps, the UI can display the steps the agent is taking to the user.
previously, running the UI and using the search agent required clone the repository and running chainlit manually

now, the core UI has been containerised for ease of sharing and testing
@VJonah VJonah requested a review from j-thomas-ucl December 15, 2025 13:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant