Xingyao Wang 0380070e98
Abstraction that allows us to develop different agents, frontend, backend, and evaluation in parallel (#68)
* move agent to langchains_agent

* remove old .env

* remove the old agent folder

* add preliminary version of Agent abstraction

* add preliminary version of the main.py

* merge controlloop and main into a Agent class

* add init

* fix json import

* fix missing arg

* get langchains_agent working after abstraction

* rename `research` to `agenthub`

* rename: rename research to agenthub

---------

Co-authored-by: huybery <huybery@gmail.com>
2024-03-20 15:09:29 -04:00

478 B

TODO

There's a lot of low-hanging fruit for this agent:

  • Strip <script>, <style>, and other non-text tags from the HTML before sending it to the LLM
  • Keep track of the working directory when the agent uses cd
  • Improve memory condensing--condense earlier memories more aggressively
  • Limit the time that run can wait (in case agent runs an interactive command and it's hanging)
  • Figure out how to run background processes, e.g. node server.js to start a server