fix: allow null function.name in streaming tool calls for OpenAI-compatible providers - #26433
Closed
xxiaoxiong wants to merge 3 commits into
Closed
Conversation
The SessionPromptData type was missing the 'format' field that exists in the OpenAPI spec. This field allows users to specify structured output formats (text or json_schema) when sending prompts. Added missing types: - OutputFormatText: for plain text output - OutputFormatJsonSchema: for JSON schema-based structured output - OutputFormat: union type of the above - format?: OutputFormat field in SessionPromptData.body This allows TypeScript users to use the format parameter without type errors, matching the actual API behavior. Fixes #26408
GitHub Actions bot comments were using the old hardcoded URL format 'https://opencode.ai/s/<id>' which now 404s. The canonical share URL format is now 'https://opncd.ai/share/<id>' and is returned by the share API. Changes: - Capture the full share URL from sessionShare.share() result - Use the returned URL directly in comments and links - Extract share ID from URL for social card image generation - Update both packages/opencode/src/cli/cmd/github.ts and github/index.ts Before: shareId = session.id.slice(-8) link = `${shareBaseUrl}/s/${shareId}` After: shareUrl = result.url // e.g. 'https://opncd.ai/share/xxx' link = shareUrl Fixes #26417
…atible providers Per OpenAI streaming spec, function.name may be null in non-first chunks. Some OpenAI-compatible providers (vLLM, LM Studio, llama.cpp server) send name: null even in the first chunk, causing 'Expected function.name to be a string' errors. Changes: - Remove strict function.name validation on first chunk - Allow creating tool calls with empty name - Update name when it arrives in later chunks - Send tool-input-start event only when name is available This makes OpenCode compatible with more OpenAI-compatible providers while maintaining correct behavior for providers that follow the spec strictly. Fixes #26412
github-actions
Bot
added
the
needs:compliance
Contributor
|
This PR doesn't fully meet our contributing guidelines and PR template. What needs to be fixed:
Please edit this PR description to address the above within 2 hours, or it will be automatically closed. If you believe this was flagged incorrectly, please let a maintainer know. |
This was referenced May 9, 2026
Contributor
|
This pull request has been automatically closed because it was not updated to meet our contributing guidelines within the 2-hour window. Feel free to open a new pull request that follows our guidelines. |
github-actions
Bot
removed
the
needs:compliance
|
I would be very grateful for this to be re-opened and merged as it fixes a very important problem with currently 12 open parallel issues. |
|
This bug prevents me from using any local agents (Qwen3.6) via LM Studio on Windows 11, so I'd be grateful that this is re-opened and merged. Thanks in advance! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Fixes #26412
This PR resolves the
Expected 'function.name' to be a stringerror when using custom OpenAI-compatible providers (vLLM, LM Studio, llama.cpp server, etc.) that sendfunction.name: nullin streaming tool call chunks.Problem
Per the OpenAI streaming API spec,
function.nameis only required in the first chunk of a tool call; subsequent chunks may sendname: null. However, some OpenAI-compatible providers sendname: nulleven in the first chunk, then send the actual name in later chunks.The current code strictly validates that
function.namemust be present when creating a new tool call (first chunk for a given index), causing all tool calls to fail with these providers.Solution
This PR makes the validation more lenient to match the behavior of the OpenAI Python SDK and other clients:
function.nameisnullin the first chunktool-input-startevent whennameis availablenamearrives in a later chunk, update the tool call and send thetool-input-startevent at that pointnamebefore completion (checked at line 623)Changes
Modified file:
packages/opencode/src/provider/sdk/copilot/chat/openai-compatible-chat-language-model.tsfunction.namevalidation, replaced with conditionaltool-input-startemissionname: toolCallDelta.function.nametoname: toolCallDelta.function?.name ?? ""namewhen it arrives in later chunksTesting
This fix enables OpenCode to work with:
The change is backward-compatible: providers that send
namein the first chunk (OpenAI, Anthropic via compatibility layer, etc.) continue to work as before.Related Issues
indexfield validation (fixed by upgrading@ai-sdk/openai-compatible)function.namevalidation at the OpenCode layer