You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If only using the CPU or an Nvidia GPU, you can run the application exclusively with Docker.
67
+
68
+
- Download the `docker-compose.yml` and `Dockerfile` (for CPU-only) or `Dockerfile-gpu` (for GPU) files
69
+
- In `docker-compose.yml`, edit the HF_TOKEN to your API token
70
+
- There are four elements which need to exist outside of the container for persistence, portability, and personalization:
71
+
- **`corpora/` directory**: this is where processed corpora (text files and vector database dumps) are saved. Change the `<local corpora directory>` line in `docker-compose.yml` to your local path for these files.
72
+
- **`metadata/` directory**: this is where various information like database credentials, user list, llm list, etc. are stored. Change the `<local metadata directory>` line in `docker-compose.yml` to your local path for these files. The elements to be manually checked and changed are `db_creds.csv`, the `app_title`, `author_name`, and `author_email` columns, `llm_list.csv`, and `user_list.csv`.
73
+
- **`models/` directory**: this is where the actual LLMs are stored. Change the `<local models directory>` line in `docker-compose.yml` to your local path for these files.
74
+
- **`secrets.toml` file**: this is where you can change the application's password. Change the `<local secrets.toml file path>` line in `docker-compose.yml` to your local path for this file.
75
+
- If you are using the CPU, delete or comment out the `deploy:` section in `docker-compose.yml`, and change the `dockerfile: Dockerfile-gpu` line to `dockerfile: Dockerfile`.
76
+
- Navigate to the directory where you saved the `docker-compose.yml` file and Dockerfile and run `docker compose up`
77
+
- The application will now be available on port 8502 by default.
78
+
79
+
### Apple silicon
80
+
If you are using Apple silicon, you won't be able to run everything in Docker because of the lack of MPS drivers. You can still use the pgvector image however.
81
+
82
+
- follow the instructions to install [local\_rag\_llm](https://github.com/dhopp1/local_rag_llm/) and [nlp\_pipeline](https://github.com/dhopp1/nlp_pipeline) individually
83
+
- Download the `docker-compose.yml` file
84
+
- From the `docker-compose.yml` file, delete the `streamlit:` line and everything below it
85
+
- Start the postgres container with `docker compose up`
86
+
- Edit your `metadata/db_creds.csv` file and change the `host` column from `localhost` to `postgres` and `username` to `postgres`
0 commit comments