feat(cohere): add support from Reasoning/Vision/Translate models #8597

maxbrunet · 2025-11-04T23:28:27Z

Description

[ What changed? Feel free to be brief. ]

Add support for reasoning in Cohere provider
Add reasoning/vision models
Add translate models (I was a bit hesitant, but I thought working on i18n of an app could be a good use case, and there is no particular extra work to support it)
Add basic tests for Cohere provider

AI Code Review

Team members only: AI review runs automatically when PR is opened or marked ready for review
Team members can also trigger a review by commenting @continue-review

Checklist

I've read the contributing guide
The relevant docs, if any, have been updated or created
The relevant tests, if any, have been updated or created

Screen recording or screenshot

[ When applicable, please include a short screen recording or screenshot - this makes it much easier for us as contributors to review and understand your changes. See this PR as a good example. ]

Tests

[ What tests were added or updated to ensure the changes work as expected? ]

Update toolSupport tests for vision model
Add basic tests for Cohere provider

Summary by cubic

Add Cohere Reasoning, Vision, and Translate model support across the provider, UI, and docs. Enables thinking traces, image input, and new models with correct tool-calling behavior and tests.

New Features
- Reasoning: handle thinking content (stream/non-stream), configurable token_budget, and autodetect for command-a-reasoning models.
- Vision: accept image_url inputs; disable tool use for command-a-vision models.
- Translate: add command-a-translate model.
- Message shaping: merge thinking with assistant messages and support tool_calls correctly.
- Ecosystem updates: VS Code schema, GUI model picker, llm-info metadata, and docs capability matrix.
- Tests: add comprehensive Cohere provider tests and tool support tests.

^{Written for commit b7f0f42. Summary will update automatically on new commits.}

cubic-dev-ai

1 issue found across 9 files

Prompt for AI agents (all 1 issues)


Understand the root cause of the following 1 issues and fix them.


<file name="core/llm/llms/Cohere.ts">

<violation number="1" location="core/llm/llms/Cohere.ts:83">
Wrapping assistant content in a `{type: &quot;text&quot;}` block assigns arrays of `MessagePart`s to the `text` field, corrupting structured assistant replies when they are already multi-part.</violation>
</file>

Since this is your first cubic review, here's how it works:

cubic automatically reviews your code and comments on bugs and improvements
Teach cubic by replying to its comments. cubic learns from your replies and gets better over time
Ask questions if you need clarification on any suggestion

_{React with 👍 or 👎 to teach cubic. Mention @cubic-dev-ai to give feedback, ask questions, or re-run the review.}

cubic-dev-ai · 2025-11-04T23:38:58Z

core/llm/llms/Cohere.ts

+          } else {
+            msg.content.push({
+              type: "text",
+              text: m.content,


Wrapping assistant content in a {type: "text"} block assigns arrays of MessageParts to the text field, corrupting structured assistant replies when they are already multi-part.

Prompt for AI agents

Address the following comment on core/llm/llms/Cohere.ts at line 83: <comment>Wrapping assistant content in a `{type: "text"}` block assigns arrays of `MessagePart`s to the `text` field, corrupting structured assistant replies when they are already multi-part.</comment> <file context> @@ -48,36 +47,44 @@ class Cohere extends BaseLLM { + } else { + msg.content.push({ + type: "text", + text: m.content, }); - lastToolPlan = undefined; </file context>

✅ Addressed in 0579434

@maxbrunet could you check this feedback?

I'm not sure that was a real problem, the provider never creates blocks from response, but addressed it

Thanks for the feedback! I've saved this as a new learning to improve future reviews.

RomneyDa

@maxbrunet appreciate this contribution and thanks for the tests.
Pretty much looks good to me, could you check on the comments briefly including the cubic one? Then should be good to merge

RomneyDa · 2025-11-07T18:30:58Z

packages/llm-info/src/providers/cohere.ts

+    {
+      model: "command-a-translate-08-2025",
+      displayName: "Command A Translate 08-2025",
+      contextLength: 8000,


probably 8192, not critical

It is documented as 8k and I wanted to be consistent with the others (128k, 256k). In all cases, there is always a little bit of wiggle room for the system prompt and for a minimal completion response

RomneyDa · 2025-11-07T18:32:54Z

core/llm/llms/Cohere.ts

      switch (value.type) {
        // https://docs.cohere.com/v2/docs/streaming#content-delta
        case "content-delta":
+          if (value.delta.message.content.thinking) {


@maxbrunet just to double check can content deltas also have text that would be lost here? or is it thinking OR text?

Yes, it is OR

https://docs.cohere.com/reference/chat#response.body.message.content

RomneyDa · 2025-11-07T18:33:26Z

core/llm/llms/Cohere.ts

+          } else {
+            msg.content.push({
+              type: "text",
+              text: m.content,


@maxbrunet could you check this feedback?

maxbrunet

Thank you @RomneyDa for the review!

maxbrunet · 2025-11-11T17:56:58Z

core/llm/llms/Cohere.ts

+          } else {
+            msg.content.push({
+              type: "text",
+              text: m.content,


I'm not sure that was a real problem, the provider never creates blocks from response, but addressed it

maxbrunet · 2025-11-11T17:57:39Z

core/llm/llms/Cohere.ts

      switch (value.type) {
        // https://docs.cohere.com/v2/docs/streaming#content-delta
        case "content-delta":
+          if (value.delta.message.content.thinking) {


Yes, it is OR

https://docs.cohere.com/reference/chat#response.body.message.content

maxbrunet · 2025-11-11T18:00:16Z

packages/llm-info/src/providers/cohere.ts

+    {
+      model: "command-a-translate-08-2025",
+      displayName: "Command A Translate 08-2025",
+      contextLength: 8000,


It is documented as 8k and I wanted to be consistent with the others (128k, 256k). In all cases, there is always a little bit of wiggle room for the system prompt and for a minimal completion response

cubic-dev-ai

1 issue found across 1 file (reviewed changes from recent commits).

1 issue found across 1 file

Prompt for AI agents (all 1 issues)


Understand the root cause of the following 1 issues and fix them.


<file name="core/llm/llms/Cohere.ts">

<violation number="1" location="core/llm/llms/Cohere.ts:87">
Overwriting `msg.content` here discards any previously merged thinking segment when we reuse the prior message, so the reasoning trace is lost. Append the new parts instead of replacing the array to keep earlier content intact.</violation>
</file>

_{React with 👍 or 👎 to teach cubic. Mention @cubic-dev-ai to give feedback, ask questions, or re-run the review.}

core/llm/llms/Cohere.ts

feat(cohere): add support from Reasoning/Vision/Translate models

9525ea8

maxbrunet requested a review from a team as a code owner November 4, 2025 23:28

maxbrunet requested review from RomneyDa and removed request for a team November 4, 2025 23:28

github-project-automation bot added this to Issues and PRs Nov 4, 2025

github-project-automation bot moved this to Todo in Issues and PRs Nov 4, 2025

dosubot bot added the size:XXL This PR changes 1000+ lines, ignoring generated files. label Nov 4, 2025

cubic-dev-ai bot reviewed Nov 4, 2025

View reviewed changes

RomneyDa reviewed Nov 7, 2025

View reviewed changes

maxbrunet commented Nov 11, 2025

View reviewed changes

cubic-dev-ai bot reviewed Nov 11, 2025

View reviewed changes

core/llm/llms/Cohere.ts Outdated Show resolved Hide resolved

Handle potential case where assistant content is already a block

b7f0f42

maxbrunet force-pushed the feat/cohere-2025-2 branch from 0579434 to b7f0f42 Compare November 11, 2025 18:14

feat(cohere): add support from Reasoning/Vision/Translate models #8597

Are you sure you want to change the base?

feat(cohere): add support from Reasoning/Vision/Translate models #8597

Uh oh!

Conversation

maxbrunet commented Nov 4, 2025 • edited by cubic-dev-ai bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

AI Code Review

Checklist

Screen recording or screenshot

Tests

Summary by cubic

Uh oh!

cubic-dev-ai bot left a comment

Choose a reason for hiding this comment

Uh oh!

cubic-dev-ai bot Nov 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cubic-dev-ai bot Nov 11, 2025

Choose a reason for hiding this comment

Uh oh!

RomneyDa left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

maxbrunet left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cubic-dev-ai bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

maxbrunet commented Nov 4, 2025 •

edited by cubic-dev-ai bot

Loading

cubic-dev-ai bot Nov 4, 2025 •

edited

Loading