wip: Integration #23

rti · 2024-02-01T22:54:54Z

TODO

test full dataset excellent articles Inacceptable performance on excellent_acticles full #39

gswikichat/vector_store_interface.py

requirements.txt

gswikichat/vector_store_interface.py

gswikichat/api.py

README.md

Dockerfile

exowanderer

I made a lot of comments and suggestions. I promise that they are 95% "PEP8 this" and "format suggestion that".

I will make a new branch with a PR that implements all of my suggestions so that neither of you has to go through them one by one.

The comment list here will be my check list after both of you approve the PR from my integrations_pep8 into the integration branch.

Dockerfile

gswikichat/api.py

gswikichat/vector_store_interface.py

Dockerfile

gswikichat/api.py

gswikichat/llm_config.py

exowanderer · 2024-02-09T21:06:29Z

I submited a PR(#24) from the integration_pep8 branch to integration branch.

This PR addresses 95% of my comments and suggestions. They are largely related to PEP8 formatting, my suggest formatting, and creating a logger.py file instead of using print statements

If you review and agree with the PR, as well as merge it, then we can close this review comments.

…py inputs; tests timed out

…i.py included; logger in api.py functional

…ed them in llm_config.py

…moved rag_piipeline from api.py; introduced rag_pipeline from rag.py into api.py

To prevent huge redownloads on llm change.

closes #35

closes #34

rti added 12 commits February 1, 2024 22:35

feat: cache huggingface models

ff3fe68

fix: sentence_transformers version

38a3bf9

chore: remove custom model based on modelfile

3fb6fd0

fix(frontend): do not filter by score for now TBD

a4c7294

chore: remove debug/test code

d38c5f0

fix: required sentence_transformers version was actually > 2.2.0

dc4501a

docs: add notes about embedding models to readme

42cdcc5

chore: add debug output to api.py

13bc12e

fix: question in prompt

4933a9a

chore: top_k 3 results for now

b23833b

wip: embeddings cache

da1017b

feat: document splitter

41ff046

rti requested review from andrewtavis and exowanderer February 2, 2024 07:19

Update .dockerignore

4e69697

exowanderer requested changes Feb 2, 2024

View reviewed changes

rti and others added 12 commits February 4, 2024 14:17

Merge branch 'main' into integration

10103c6

docs: note on how to dev locally

0ee6ed5

docs: add research_log.md

7a2c955

feat: set top_k via api

0a5e2be

feat: support en and de on the api to switch prompts

332e3dc

feat: cache embedding model during docker build

6225fcc

wip: smaller chunk size, 5 sentences for now

4877807

chore: remove comment

da9859d

feat: enable embeddings cache (for developmnet)

291aaaf

feat: add document cleaner

936d83e

Merge branch 'main' into integration

1b88437

docs: long docker run options

3e0b8f4

rti requested review from exowanderer February 9, 2024 09:38

rti added 5 commits February 9, 2024 09:40

fix: access mode

edf5eb2

fix: redraw loading animation on subsequent searches

63baf2b

wip: workaround for runpod.io http port forwarding

56a7b8c

feat: switch to openchat 7b model

8e05473

Merge branch 'openchat' into integration

8276e35

exowanderer requested changes Feb 9, 2024

View reviewed changes

Dockerfile Outdated Show resolved Hide resolved

gswikichat/api.py Outdated Show resolved Hide resolved

gswikichat/llm_config.py Outdated Show resolved Hide resolved

exowanderer and others added 13 commits February 11, 2024 12:20

added logging via logger with Handler to api.py; PEP8 formatted api.py

22b04d0

debugging use of homepage instead of hard coded endpoint values

10f6b21

returning to previous to restart without errors

bfbd245

renewed app.mount; bug fixed PEP8 changes in api.py; reformatted rag.…

7b6ba0a

…py inputs; tests timed out

returned to stablelm2 model for testing purposes. PEP8 upgrades in ap…

0428f87

…i.py included; logger in api.py functional

added OLLAMA_MODEL_NAME and OLLAMA_URL as environment variables; call…

8104dde

…ed them in llm_config.py

created logger.py to serve get_logger to all modules

fbc4591

created a rag_pipeline in the rag.py based on the usage in api.py; re…

caecfd1

…moved rag_piipeline from api.py; introduced rag_pipeline from rag.py into api.py

UPdated with PEP8 formatting in vector_store_interface.py

5c0b4d0

chore(Dockerfile): install python deps early

8833af7

To prevent huge redownloads on llm change.

fix(sentence-transformers): use cuda if available

9ee8a32

fix(frontend): run from webserver root

b2357e3

feat: store embedding cache in volume

b518abf

rti mentioned this pull request Feb 12, 2024

Embedding generation runs on CPU only #28

Closed

2 tasks

rti added 2 commits February 12, 2024 22:30

feat(start.sh): pull llm using ollama (if not built into container)

69800b0

closes #35

feat(ollama): use chat api to leverage prompt templates

7803649

closes #34

This was referenced Feb 12, 2024

Support overwriting model via environment #35

Closed

Use ollama chat-generation #34

Closed

docs: fix run cmd

ff1fcab

rti mentioned this pull request Feb 19, 2024

feat: switch to openchat 7b model #25

Closed

rti merged commit 4100064 into main Feb 27, 2024

rti deleted the integration branch February 27, 2024 15:12

wip: Integration #23

wip: Integration #23

Uh oh!

Conversation

rti commented Feb 1, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

TODO

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

exowanderer left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

exowanderer commented Feb 9, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

rti commented Feb 1, 2024 •

edited

Loading