Fix options being passed to the ollama chat api #179
Conversation
|
Hi @alappe! I assume this is still a needed change, right? The doc links you provided sure make it appear that way. (Thanks for those, btw!) @mustela, your recent note here leads me to believe you're successfully using the library with Ollama. Are you doing it with or without this change? I'm trying to understand if making this change breaks backward compatibility or if this is just how it must be done going forward. Thanks! |
|
Hi @brainlid, yes, this is still needed. It's still possible to use the library with ollama at the moment, because if you're not providing the options (in the right place), the defaults apply. But you cannot change options without the change. IMHO it breaks nothing because the current state just didn't work right with ollama. |
alappe
force-pushed
the
ollama_fix_options
branch
from
6ad8e91 to
471951c
Compare
|
@brainlid I rebased the PR… |
|
Thanks @alappe! I'm trusting that this is working well as I'm not actively testing Ollama. I appreciate the help improving the library! |
* main: Support for file with file_id in ChatOpenAI (#283) Fix options being passed to the ollama chat api (#179) Support for json_response in ChatModels.ChatGoogleAI (#277) support streaming responses from mistral (#287) feat: File urls for Google (#286) added ChatPerplexity to the docs - updated ContentPart docs
* main: Support for file with file_id in ChatOpenAI (#283) Fix options being passed to the ollama chat api (#179) Support for json_response in ChatModels.ChatGoogleAI (#277) support streaming responses from mistral (#287) feat: File urls for Google (#286) added ChatPerplexity to the docs - updated ContentPart docs
* main: Support for file with file_id in ChatOpenAI (#283) Fix options being passed to the ollama chat api (#179) Support for json_response in ChatModels.ChatGoogleAI (#277) support streaming responses from mistral (#287) feat: File urls for Google (#286) added ChatPerplexity to the docs - updated ContentPart docs
* main: updated readme for version install updated mix.exs version for release updated changelog for v0.3.3 added content part description to OpenAI module doc for file uploads fixed doc typo adds telemetry (#284) check that the requested tool_name exists - return an error if it does not exist in the chain added LLMChain.run_until_tool_used/3 (#292) Support for file with file_id in ChatOpenAI (#283) Fix options being passed to the ollama chat api (#179) Support for json_response in ChatModels.ChatGoogleAI (#277) support streaming responses from mistral (#287) feat: File urls for Google (#286)
2 participants
I noticed that https://github.com/ollama/ollama/blob/main/docs/api.md#generate-a-chat-completion and https://github.com/ollama/ollama/blob/main/docs/modelfile.md#valid-parameters-and-values document that those options all need to be passed to the
options: {}sub structure.My local instance testing with llama
v3.1and ollamav0.3.10complains becausenum_gqais unknown. I left it in there for now because other models might need it.