Open andre-scheinwald opened 3 months ago
@andre-scheinwald
Hmm, I suspect that you're using the volume created by the .exe
launcher. Could you check if you're using the same volume ? if so, could you delete the containers & volume all together and try again? since you're going to use docker-compose yourself.
@wwwy3y3 Deleting the containers and volumes got me through the connection steps, thanks!
Basically what I'm trying to do is have the docker version connect to one database. And the launcher connect to a different database. That way I can retain work in both.
Now that I have the connections and relationships working in the docker and launcher versions. I've tried asking questions in both. The launcher version works fine. However the docker version returns a "Failed to create asking task" error. I haven't tried running them simultaneously when I get this error message in the docker version. And I ensured that the launcher container is stopped.
Do you have any thoughts on this?
The launcher version works fine. However the docker version returns a "Failed to create asking task" error. I haven't tried running them simultaneously when I get this error message in the docker version. And I ensured that the launcher container is stopped.
I'm guessing you're putting an invalid OpenAI API key in the docker .env
file. You could copy the .env
file launcher's been using at ~/.wrenai/.env
.
would be sth like following:
COMPOSE_PROJECT_NAME=wren
PLATFORM=linux/amd64
# service port
WREN_ENGINE_PORT=8080
WREN_ENGINE_SQL_PORT=7432
WREN_AI_SERVICE_PORT=5555
# version
# CHANGE THIS TO THE LATEST VERSION
WREN_PRODUCT_VERSION=0.3.5
WREN_ENGINE_VERSION=0.4.5
WREN_AI_SERVICE_VERSION=0.4.0
WREN_UI_VERSION=0.5.7
WREN_BOOTSTRAP_VERSION=0.1.4
# keys
# CHANGE THIS TO YOUR OPENAI API KEY
OPENAI_API_KEY=<API KEY> # -> replace with your own key
OPENAI_GENERATION_MODEL=gpt-4o
# SQL Protocol
PG_USERNAME=wren-user
# PG_PASSWORD will be replaced with random string by launcher
PG_PASSWORD=...
# user id (uuid v4)
USER_UUID=f5bc7ee1-0a0b-4ebe-b02f-533d7bb5421a
# for other services
POSTHOG_API_KEY=...
POSTHOG_HOST=https://app.posthog.com
TELEMETRY_ENABLED=true
# the port exposes to the host
# OPTIONAL: change the port if you have a conflict
HOST_PORT=3001
AI_SERVICE_FORWARD_PORT=5556
Related discussion on discord: https://discord.com/channels/1227143286951514152/1243764365887537162
I wish it was the openapi key! I found two different issues.
Failed to prepare semantics: Error code: 400 - {'error': {'message': "This model's maximum context length is 8192 tokens, however you requested 10018 tokens (10018 in your prompt; 0 for the completion). Please reduce your prompt; or completion length.", 'type': 'invalid_request_error', 'param': None, 'code': None}}
I'm only loading in 16 tables but the column length in a number of these tables is over 100. Decreasing the number of tables used resolves the token error. It appears that I can't delete tables from the modeling tab and deploy changes. I have to reset under settings and start over.
I wish it was the openapi key! I found two different issues.
- One is there appears to be a limit to number of tables + columns? I finally noticed that when asking a question not only would I get the "Failed to create asking task error message" but the terminal was printing the following error message:
Failed to prepare semantics: Error code: 400 - {'error': {'message': "This model's maximum context length is 8192 tokens, however you requested 10018 tokens (10018 in your prompt; 0 for the completion). Please reduce your prompt; or completion length.", 'type': 'invalid_request_error', 'param': None, 'code': None}}
I'm only loading in 16 tables but the column length in a number of these tables is over 100. Decreasing the number of tables used resolves the token error. It appears that I can't delete tables from the modeling tab and deploy changes. I have to reset under settings and start over.
- The next error was related to invalid model. I thought this had worked in the past - but maybe not. In my .env.local I had written "gpt-4.0-turbo". I changed this to "gpt-4-turbo" and can now retrieve results. Would it be possible to add the text for the model options to the .env.example file?
@andre-scheinwald
thanks for raising the issue. "It appears that I can't delete tables from the modeling tab and deploy changes. I have to reset under settings and start over." -> do you mean deployment was failed, so you couldn't make changes?
thanks for your suggestions!
Describe the bug I'm trying to run the tool using the github repo based on these instructions under "How to start". I've created and edited my .env.local file to run off port 8080. 3000 is occupied by the .exe launcher. I've ensured that the docker container the .exe creates and runs is stopped. When I run
docker-compose --env-file .env.local up
I navigate to localhost:8080 and can select db type. On the next screen for connection set up I put in all my credentials and select Next. That is when I encounter the "Internal server error...returning * - table project has no column named configurations" error message.I've tested this on two different postgres databases and get the same error message.
To Reproduce Steps to reproduce the behavior:
docker-compose --env-file .env.local up
.Expected behavior The data source to properly connect
Screenshots![image](https://github.com/Canner/WrenAI/assets/130578303/f1f19a13-004c-4587-9a37-c43854d58441)
Desktop (please complete the following information):
WrenAI Information
Additional Information I don't have a project table in my db. And considering this is insert - I'm assuming this is within wrenai?