Closed timucindusunur closed 5 months ago
but it gives an error during installation.
What's the error? Please fill in the bug template when creating issues.
(OpenDevin) root@DESKTOP-1TC1K9D:/home/timuc/OpenDevin# make run Running the app... /bin/sh: 1: [: unexpected operator make[1]: Entering directory '/home/timuc/OpenDevin' Starting backend... make[1]: Entering directory '/home/timuc/OpenDevin' Starting frontend...
opendevin-frontend@0.1.0 start vite
VITE v5.2.8 ready in 641 ms
➜ Local: http://localhost:3001/ ➜ Network: use --host to expose INFO: Started server process [51464] INFO: Waiting for application startup. INFO: Application startup complete. INFO: Uvicorn running on http://127.0.0.1:3000 (Press CTRL+C to quit) INFO: 127.0.0.1:57184 - "GET /litellm-models HTTP/1.1" 200 OK INFO: 127.0.0.1:57188 - "GET /litellm-agents HTTP/1.1" 200 OK INFO: 127.0.0.1:57202 - "GET /litellm-models HTTP/1.1" 200 OK INFO: 127.0.0.1:57208 - "GET /litellm-agents HTTP/1.1" 200 OK INFO: ('127.0.0.1', 57220) - "WebSocket /ws" [accepted] INFO: connection open
============== STEP 0
PLAN: bana mali danışmanlık firması için web sitesi tasarla ve çalıştır. Detaylı ve kapsamlı bir çalışma olsun.
INFO: HINT: You're not currently working on any tasks. Your next action MUST be to mark a task as in_progress.
Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new LiteLLM.Info: If you need to debug this error, use `litellm.set_verbose=True'.
Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new LiteLLM.Info: If you need to debug this error, use `litellm.set_verbose=True'.
Queries made here for "http://localhost:3001" crash and do not return any results. There is no movement other than the background output I gave above.
I can repro this issue.
steps:
Set Env vars:
LLM_MODEL="gemini/gemini-pro"
LLM_API_KEY="A....."
Run
PYTHONPATH="./" python opendevin/main.py -d ./workspace/ -i 100 -t "Write a bash script that prints hello world"
This works, with output of the form
Running agent MonologueAgent (model: gemini/gemini-pro, directory: ./workspace/) with task: "Write a bash script that prints hello world"
==============
STEP 0
PLAN:
Write a bash script that prints hello world
ACTION:
CmdRunAction(command='echo "hello world"', background=False, action='run')
OBSERVATION:
hello world
==============
STEP 1
...
Then start server with make run
and put Write a bash script that prints hello world
into the prompt.
results in, among other logs
litellm.exceptions.RateLimitError: VertexAIException - vertexai import failed please run `pip install google-cloud-aiplatform`
I ran pip install google-cloud-aiplatform
and restarted the server, got a different error
VertexAIException - Your default credentials were not found. To set up Application Default Credentials, see https://cloud.google.com/docs/authentication/external/set-up-adc for more information.
My best guess is LLM_MODEL="gemini/gemini-pro"
which the command line invocation uses, is not the same as gemini-pro
set from the UI.
https://github.com/OpenDevin/OpenDevin/pull/654 fixes this issue.
Well, merging 654, and selecting gemini/gemini-pro
from the model drop down box as opposed to gemini-pro
because both entries are available.
@rbren would you mind merging #654 when you have a chance please.
@foragerr You are exactly right about there being an issue gemini/gemini-pro vs gemini-pro. See my comments here https://github.com/OpenDevin/OpenDevin/issues/653#issuecomment-2035536785 and following.
@jpshack-at-palomar gemini/gemini-pro
should be available from the UI dropdown now.
Going to close this one, but we definitely need some better model management. Thanks all!
So basically you can check the Pre-requisites section from litellm documentation now Gemini Pro is working 😁
PYTHONPATH="./" python opendevin/main.py -d ./workspace/ -i 100 -t "Write a bash script that prints hello world"
The command line is working with Gemini AI, but it gives an error during installation. What could be the reason for this?
config.toml
LLM_MODEL="gemini/gemini-pro" LLM_API_KEY="A....." WORKSPACE_DIR="./workspace"