Code

Ollama, Mistral AI, etc.#7296

Code

Merged

williamstein merged 45 commits intomastersagemathinc/cocalc:masterfrom support-ollama-apisagemathinc/cocalc:support-ollama-apiCopy head branch name to clipboardMar 17, 2024

+3,690-1,684

Commits

Commits on Feb 23, 2024

• server/llm: first steps towards supporting ollama

  haraldschilly

  committed
  fd4a48e

• ollama: starting with configuration + frontend

  haraldschilly

  committed
  b6f4c87

• util/llm: refactoring to keep my sanity (and fixing circular imports)

  haraldschilly

  committed
  71140d3

• server/ollama: some progress ...

  haraldschilly

  committed
  3413932

Commits on Feb 26, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  7c22214

• frontend/ollama: various fixes, logo, starting chat integration (+bugfixes?) etc.

  haraldschilly

  committed
  d6c763e

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  ed19f8c

Commits on Feb 27, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  e06be12

• llm: make model selection a dropdown, show more of the explanations, add ollama description, and fix bug with ai-formula (no project context)

  haraldschilly

  committed
  35abc06

Commits on Feb 28, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  d8ffc3c

• frontend/jupyter: cleanup weird "chatgpt" object, fix codebar offset, box around cell number, etc

  haraldschilly

  committed
  a1ddbed

Commits on Mar 8, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  daba4cc

• llm: get rid of silly "string" fallback typing

  haraldschilly

  committed
  eef5a85

Commits on Mar 11, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  dcfcd94

• llm: start supporting mistral ai

  haraldschilly

  committed
  ab08bfa

Commits on Mar 12, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  493656c

• llm: adding mistral and more in-depth refactoring

  haraldschilly

  committed
  472a644

• frontend/latex/llm: improve latex formula postprocessing

  haraldschilly

  committed
  e8509d5

• frontend/llm: more towards mistral

  haraldschilly

  committed
  2b0861c

• frontend/tex/ai formula: further fixes and tweaks

  haraldschilly

  committed
  09613a6

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  ce9062e

• frontend/latex/ai formula: further debug dialog while testing LLM

  haraldschilly

  committed
  94f5394

Commits on Mar 13, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  3a2ac62

• frontend/chat/llm: untangling ollama-[model] and other model names -- to be fixed later

  haraldschilly

  committed
  2c507e6

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  141ccc3

Commits on Mar 14, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  4dc311c

• npm: update langchain

  haraldschilly

  committed
  fabb771

• llm: do not cache ollama, tweak model selector, count tokens, etc.

  haraldschilly

  committed
  91821ab

• frontend: refactor more chatgpt → llm

  haraldschilly

  committed
  f0d64e7

• llm: dynamic standard system prompt – otherwise mistral too confused

  haraldschilly

  committed
  1e51d9b

Commits on Mar 15, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  44aa6ee

• llm: add support for OpenAI GPT-4 Turbo

  haraldschilly

  committed
  3c706bf

• llm: make Mistral Large a paid model

  haraldschilly

  committed
  99285c9

• frontend/account/other-settings: convert to functional component, and use the ModelSwitch component

  haraldschilly

  committed
  44f7b01

• frontend/latex/ai formula: add button to insert full reply, since sometimes it is interesting

  haraldschilly

  committed
  73b40f6

• frontend/llm: fixing #7341 by being specific about which LLM tools to enable/disable in course projects

  haraldschilly

  committed
  d9059ed

• util/llm: minor fixes and tweaks

  haraldschilly

  committed
  39a378a

• frontend/llm: respect AI student project limit in AI formula generator; throw proper error when querying client/llm without an enabled LLM for a specific tag

  haraldschilly

  committed
  d3150c5

• frontend/llm: further debug and explain what is going on with student course project specific limitations

  haraldschilly

  committed
  5d3f293

• frontend/llm: collect LLM descriptions in the universal llm-util

  haraldschilly

  committed
  cb66d65

• server/llm: fix token counter for mistral and ollama

  haraldschilly

  committed
  6b3ec8f

• server/llm: drop Palm2 support

  haraldschilly

  committed
  e4ed26c

Commits on Mar 17, 2024

• Merge remote-tracking branch 'origin/master' into support-ollama-api

  haraldschilly

  committed
  00834d1

• npm: fork mistralai client, override its dependency, document why, update langchain

  haraldschilly

  committed
  1e9def5

• llm: add a 8k limited GPT4 Turbo variant, to avoid committing to too much money

  haraldschilly

  committed
  28b989a