feat: Complete DeepSeek API adapter implementation#11130
Open
dixoxib wants to merge 44 commits intocontinuedev:mainfrom
Open
feat: Complete DeepSeek API adapter implementation#11130dixoxib wants to merge 44 commits intocontinuedev:mainfrom
dixoxib wants to merge 44 commits intocontinuedev:mainfrom
Conversation
…model provider updates - Add DeepSeek model provider implementation with proper FIM support - Implement DeepSeek API adapter with OpenAI-compatible interface - Add tool call support and thinking mode integration - Update model provider configuration and onboarding - Add comprehensive type definitions and validation - Update documentation for DeepSeek model capabilities - Fix file rename from Deepseek.ts to DeepSeek.ts for consistency
… DeepSeek integration
… comprehensive tests
- Enhance DeepSeek provider with improved FIM support - Update onboarding configuration for DeepSeek - Refactor token counting and autodetection logic - Improve system tool call interception - Streamline chat response streaming - Update UI and extension components for better integration
- Refactor token counting logic - Enhance DeepSeek provider capabilities - Update chat and edit templates - Improve system tool call interception - Streamline API adapters and converters - Add unit test enhancements
Integrate latest upstream changes including: - zAI provider support - Background job service - CLI tool improvements - Updated model configurations - Maintain DeepSeek integration
- Rewrite setupProviderConfig to use direct model config instead of uses/with syntax - Add roles property to model config schema - Improve DeepSeek message conversion with toChatBody - Update conversationCompaction imports - Add roles field to config schema for VS Code - Update DeepSeek provider description - Add debug logging to DeepSeekApi adapter
- Simplify DeepSeek onboarding config to use direct model objects - Improve token counting for DeepSeek models (reserve full maxTokens) - Fix FIM support tests for DeepSeek models - Adjust maxTokens for DeepSeek Reasoner to 32k (API limit) - Update DeepSeek provider info with correct context lengths - Enhance DeepSeek converters to handle max_completion_tokens - Clean up imports and schema definitions - Update documentation with accurate DeepSeek capabilities - Remove debug console logs from DeepSeek adapter
dosubot
bot
added
the
size:XXL
Contributor
|
I have read the CLA Document and I hereby sign the CLA 1 out of 2 committers have signed the CLA. |
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
6 issues found across 35 files
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="core/config/yaml/models.ts">
<violation number="1" location="core/config/yaml/models.ts:55">
P1: Forcing `contextLength` to `DEFAULT_CONTEXT_LENGTH` in YAML model conversion can override provider runtime autodetection (e.g., Ollama), causing incorrect context windows for unknown/autodetected models.</violation>
</file>
<file name="packages/openai-adapters/src/types.ts">
<violation number="1" location="packages/openai-adapters/src/types.ts:63">
P1: `deepseek` was added to `OpenAIConfigSchema` even though `DeepseekConfigSchema` already handles that discriminator in `LLMConfigSchema`, creating a duplicate discriminator value in the same `z.discriminatedUnion`.</violation>
</file>
<file name="packages/openai-adapters/src/apis/DeepSeek.ts">
<violation number="1" location="packages/openai-adapters/src/apis/DeepSeek.ts:317">
P2: FIM completions are beta-only and require the beta base URL. Building the endpoint as `/completions` against the default base (`https://api.deepseek.com/`) no longer targets the beta API, so FIM will fail unless callers override apiBase manually.</violation>
</file>
<file name="packages/openai-adapters/src/util/deepseek-converters.ts">
<violation number="1" location="packages/openai-adapters/src/util/deepseek-converters.ts:327">
P2: `top_logprobs` is included even when `logprobs` is false/undefined, but DeepSeek requires `logprobs=true` whenever `top_logprobs` is used. This can generate invalid requests that the API rejects.</violation>
</file>
<file name="core/llm/llms/DeepSeek.ts">
<violation number="1" location="core/llm/llms/DeepSeek.ts:73">
P2: Non-streaming chat ignores top-level options.model, unlike streamChat/streamFim, causing inconsistent model selection.</violation>
<violation number="2" location="core/llm/llms/DeepSeek.ts:217">
P2: supportsPrefill() reports true unconditionally even though DeepSeek prefill is beta-only. This misreports capability on non‑beta API bases and can enable beta-only prompt paths against non‑beta endpoints.</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
- Add one-off context when rerunning by tagging
@cubic-dev-aiwith guidance or docs links (includingllms.txt) - Ask questions if you need clarification on any suggestion
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
…plicate deepseek discriminator
Author
|
I have read the CLA Document and I hereby sign the CLA |
Co-authored-by: cubic-dev-ai[bot] <191113872+cubic-dev-ai[bot]@users.noreply.github.com>
Co-authored-by: cubic-dev-ai[bot] <191113872+cubic-dev-ai[bot]@users.noreply.github.com>
…roviders - Remove trailing slash normalization in DeepSeekApi constructor - Update test expectations to match actual behavior - All tests pass, including main integration test
Contributor
There was a problem hiding this comment.
2 issues found across 2 files (changes from recent commits).
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="packages/openai-adapters/src/apis/DeepSeek.ts">
<violation number="1" location="packages/openai-adapters/src/apis/DeepSeek.ts:50">
P2: apiBase trailing-slash normalization was removed; relative URL construction can drop path segments for custom bases like https://host/v1, misrouting requests.</violation>
<violation number="2" location="packages/openai-adapters/src/apis/DeepSeek.ts:316">
P2: FIM is a beta feature that requires the beta base URL; building the endpoint as `new URL("completions", this.apiBase)` sends FIM requests to the non‑beta `/completions` path when using the default base URL, which breaks FIM. Restore beta-aware routing or force the beta base URL for FIM calls.</violation>
</file>
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
- Update test expectations to match actual regex logic - deepseek-chat matches /deepseek/ and /r1|reasoner|-chat/ regex combo - Keep deepseek-coder as non-recommended
- Fix code style issues in DeepSeek.ts and DeepSeekApi.test.ts - Ensure consistent formatting before PR merge
…licit type annotation from static defaultOptions to ensure proper inheritance\n- Remove overridden chat() method that bypassed adapter disable in tests\n- Tests now pass: default API base correctly detected and chat requests use mocked fetch
- Move applyToolOverrides import to avoid potential circular dependencies - Filter chat() accumulation to only assistant messages (ignore thinking messages) - Maintains backward compatibility with existing functionality
Contributor
There was a problem hiding this comment.
1 issue found across 2 files (changes from recent commits).
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="packages/openai-adapters/src/apis/DeepSeek.ts">
<violation number="1" location="packages/openai-adapters/src/apis/DeepSeek.ts:123">
P2: Chat endpoint base normalization is applied to all chat requests, causing `/beta/`-scoped `apiBase` values to be stripped for normal chat and potentially breaking custom `/beta/` deployments.</violation>
</file>
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
…for custom "/beta/" deployments.
Contributor
There was a problem hiding this comment.
2 issues found across 5 files (changes from recent commits).
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="packages/openai-adapters/src/apis/DeepSeek.ts">
<violation number="1" location="packages/openai-adapters/src/apis/DeepSeek.ts:123">
P1: Removed `/beta/` base normalization causes malformed or incorrect chat endpoints for clients configured with `apiBase` ending in `/beta/`.</violation>
<violation number="2" location="packages/openai-adapters/src/apis/DeepSeek.ts:321">
P1: FIM endpoint now double-prefixes `beta` when `apiBase` already includes `/beta/`, breaking beta-configured clients.</violation>
</file>
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
|
|
||
| const endpoint = new URL( | ||
| isPrefixCompletion ? "beta/chat/completions" : "chat/completions", | ||
| this.apiBase, |
Contributor
There was a problem hiding this comment.
P1: Removed /beta/ base normalization causes malformed or incorrect chat endpoints for clients configured with apiBase ending in /beta/.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At packages/openai-adapters/src/apis/DeepSeek.ts, line 123:
<comment>Removed `/beta/` base normalization causes malformed or incorrect chat endpoints for clients configured with `apiBase` ending in `/beta/`.</comment>
<file context>
@@ -120,7 +120,7 @@ export class DeepSeekApi extends OpenAIApi {
const endpoint = new URL(
isPrefixCompletion ? "beta/chat/completions" : "chat/completions",
- this.apiBase.endsWith("/beta/") ? this.apiBase.slice(0, -5) : this.apiBase,
+ this.apiBase,
);
</file context>
Suggested change
| this.apiBase, | |
| this.apiBase.endsWith("/beta/") ? this.apiBase.slice(0, -5) : this.apiBase, |
Author
There was a problem hiding this comment.
"/beta/…" is an unique prefill endpoint of the DeepSeek API. If you want to remove it for custom deployments, an apiBase provider check could be added
const apiBaseBeta = apiBase.contains("api.deepseek.com") ? apiBase + "beta/" : apiBase;
…
new URL( isPrefixCompletion ? apiBaseBeta : apiBase, …
| ): AsyncGenerator<ChatCompletionChunk, any, unknown> { | ||
| ): AsyncGenerator<ChatCompletionChunk> { | ||
| const warnings: string[] = []; | ||
| const endpoint = new URL("beta/completions", this.apiBase); |
Contributor
There was a problem hiding this comment.
P1: FIM endpoint now double-prefixes beta when apiBase already includes /beta/, breaking beta-configured clients.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At packages/openai-adapters/src/apis/DeepSeek.ts, line 321:
<comment>FIM endpoint now double-prefixes `beta` when `apiBase` already includes `/beta/`, breaking beta-configured clients.</comment>
<file context>
@@ -318,10 +318,7 @@ export class DeepSeekApi extends OpenAIApi {
- this.apiBase.endsWith("/beta/") ? "completions" : "beta/completions",
- this.apiBase,
- );
+ const endpoint = new URL("beta/completions", this.apiBase);
const deepSeekBody = convertToFimDeepSeekRequestBody(body, warnings);
</file context>
Suggested change
| const endpoint = new URL("beta/completions", this.apiBase); | |
| const endpoint = new URL( | |
| this.apiBase.endsWith("/beta/") ? "completions" : "beta/completions", | |
| this.apiBase, | |
| ); |
Author
There was a problem hiding this comment.
"/beta/completion" is an unique endpoint of the DeepSeek FIM API. If you want to remove it for custom deployments, an apiBase provider check could be added
const apiBaseBeta = apiBase.contains("api.deepseek.com") ? apiBase + "beta/" : apiBase; // or even "beta/completion" to allow custom "fim" paths
…
new URL( apiBaseBeta, …
Contributor
There was a problem hiding this comment.
1 issue found across 2 files (changes from recent commits).
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="core/llm/countTokens.ts">
<violation number="1" location="core/llm/countTokens.ts:464">
P1: Prefill token budgeting double-counts `msgsCopy` as both non-negotiable and prunable history, causing incorrect pruning and possible premature context overflow errors.</violation>
</file>
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
- Fix double-counting of tokens in prefill scenarios that caused incorrect pruning - Treat last assistant message as non-negotiable instead of entire conversation - Prevents 'Error parsing chat history: no user/tool message found' during edits - Update DeepSeek token multiplier from 1.00 to 1.05 Resolves issue where edit requests with DeepSeek and other prefill-enabled models would fail due to incorrect token budgeting in compileChatMessages.
Author
|
@cubic-dev-ai something went wrong, re-run a review. |
Contributor
@dixoxib I have started the AI code review. It will take a few minutes to complete. |
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
1 participant
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
Enhanced DeepSeek API adapter from a minimal 73‑line implementation to a complete, production‑ready adapter with all features expected from a modern LLM provider integration following official docs.
Key improvements:
/modelsendpointchat()(function seems not fully implemented yet)Files changed: 39 files, including core implementation, tests, documentation, and configuration updates, small fixes.
AI Code Review
@continue-reviewChecklist
Screen recording or screenshot
Key demonstrations:
reasoning_contentin responsessv.mp4
Tests
Added/updated tests:
packages/openai-adapters/src/test/deepseek-converters.test.ts– Comprehensive unit tests for converter functionscore/llm/llms/DeepSeek.unit.test.ts– Unit tests for DeepSeek provider classcore/llm/llms/DeepSeek.vitest.ts– Integration testspackages/openai-adapters/src/test/DeepSeekApi.test.ts– API adapter testcore/llm/llms/DeepSeek.tools.test.ts– Thinking tool chain testThe enhancement addresses limitations in the current minimal implementation and enables full DeepSeek functionality including Agent mode with tool calling for Reasoner model.
I have read the CLA Document and I hereby sign the CLA
Continue Tasks: 🔄 7 running — View all
Summary by cubic
Complete DeepSeek provider with chat, reasoning, tools, and FIM (beta) support, plus onboarding/docs and UI updates. Improves assistant-only streaming, usage/cost reporting (cache hit/miss), token counting with a prefill assistant-last fix and 1.05 multiplier, and preserves
apiBasewith correct FIM/betaendpoint handling.New Features
reasoning_content, thinking→tool call pairing, tool choice, model autodetect, and OS‑models edit template.deepseek-fim-betaonly;supportsFimtrue for FIM beta; FIM uses/beta/completionsat request time (no/betainapiBase).maxTokensfor DeepSeek, 1.05 token multiplier, improved pruning, streamed usage stats, DeepSeek pricing with cache hit/miss; fixed token counting in prefill (assistant‑last) edits to prevent errors.apiBasepreserved exactly and defaulted; addedllm-infoentries, onboarding/UI and VS Code updates;packages/openai-adaptersDeepSeek adapter, converters, and tests.Migration
provider: deepseekwith models:deepseek-chat,deepseek-reasoner,deepseek-fim-beta(for FIM).apiBasedefaults tohttps://api.deepseek.com/; do not include/beta(FIM endpoint adds it automatically). SetmaxTokens(chat: 8k, reasoner: 32k).deepseek-coderwith the newDeepSeekclass and models; provider ID remainsdeepseek.Written for commit 129334f. Summary will update on new commits.