LLamaIndex Integration

[gallegi]

• [x] LlamaIndex integration using workflow
• [x] Continue conversation (ask follow-up questions)
• [x] RAG pipeline, allow dropping multiple pdf, doc, txt files and ask related questions
• [x] Fix text cut off in the chat box
• [x] Be able to display markdown in the chat box
• [x] Manage history and retrieval context not to exceed context length
• [x] Allow configuration in setting dialog for RAG (embed model, retrieval top k, chunk size, ...) and generation (context length, temperature, top k, top p)
• [x] Preloading indicator: Add text animation on model loading / downloading or prompt reading

[vietanhdev]

Error due to missing package:

 File "/opt/homebrew/Caskroom/miniforge/base/envs/la/lib/python3.11/site-packages/llama_index/core/readers/file/base.py", line 67, in _try_loading_included_file_formats
    raise ImportError("`llama-index-readers-file` package not found")
ImportError: `llama-index-readers-file` package not found

[vietanhdev]

Error while inference:

 File "/opt/homebrew/Caskroom/miniforge/base/envs/la/lib/python3.11/site-packages/llama_cpp/llama_chat_format.py", line 289, in _convert_text_completion_chunks_to_chat
    for i, chunk in enumerate(chunks):
  File "/opt/homebrew/Caskroom/miniforge/base/envs/la/lib/python3.11/site-packages/llama_cpp/llama.py", line 1269, in _create_completion
    raise ValueError(
ValueError: Requested tokens (2387) exceed context window of 2048

File: sockets.txt

Question:

 Is socket supported in AnyLearning?

Fix:

This works when I update all context length (2048) to 4096. The answer was based on the content of the text file. Good job!

TODO: Make it configurable.

[vietanhdev]

Implementing https://github.com/nrl-ai/llama-assistant/issues/13.

[vietanhdev]

LGTM.