DeepSeek models inside Copilot Chat, with secure defaults, workspace context, and local usage visibility.
A model-provider bridge focused on code quality, accuracy, and developer productivity within your existing VS Code workflow.
DeepSeek Bridge is a VS Code language model provider extension. It adds DeepSeek models directly to Copilot Chat's model picker while keeping your normal Copilot flow (tools, agent mode, instructions, MCP, skills).
Highlights
- DeepSeek v4 Flash and DeepSeek v4 Pro in the Copilot Chat model picker.
- Thinking Effort control in the picker (
none,high,max). - Optional image handling via a configurable vision helper model.
- Workspace retrieval enabled by default for project-aware answers.
- Local usage and cost display in the status bar and details panel.
- Endpoint trust controls and guarded debug dump behavior.
Data Handling
- API keys are stored in VS Code
SecretStorageby default. - DeepSeek Bridge does not send usage analytics or chat logs to the extension publisher.
- When local usage history is enabled, it stores token/cost totals, model IDs, timestamps, and local session IDs. It does not store prompt or response text.
- Workspace retrieval sends selected local context to the configured model endpoint as part of a chat request. You can disable it from settings or the command palette.
- Full payload dumps are off by default and require explicit opt-in:
debugMode = verboseallowSensitiveRequestDumps = true
Quick Start
- Run
DeepSeek Bridge: Connect API Key. - Open Copilot Chat and select
DeepSeek v4 FlashorDeepSeek v4 Pro. - Optional: run
DeepSeek Bridge: Choose Vision Helper Model. - Optional: run
DeepSeek Bridge: Inspect Last Retrieved Contextto review the latest context sent with a request.
Commands
DeepSeek Bridge: Connect API KeyDeepSeek Bridge: Open API Key PageDeepSeek Bridge: Disconnect API KeyDeepSeek Bridge: Choose Vision Helper ModelDeepSeek Bridge: Show UsageDeepSeek Bridge: Reset Current Usage SessionDeepSeek Bridge: Enable Workspace RetrievalDeepSeek Bridge: Disable Workspace RetrievalDeepSeek Bridge: Inspect Last Retrieved ContextDeepSeek Bridge: Clear Retrieval CacheDeepSeek Bridge: Show Security StatusDeepSeek Bridge: Trust Current API EndpointDeepSeek Bridge: Reset Trusted API EndpointsDeepSeek Bridge: Show LogsDeepSeek Bridge: Open Request Dumps Folder
Settings
| Setting | Default | Purpose |
|---|---|---|
deepseek-bridge.baseUrl |
https://api.deepseek.com |
DeepSeek-compatible endpoint |
deepseek-bridge.maxTokens |
0 |
Max completion tokens (0 = API default) |
deepseek-bridge.modelIdOverrides |
official IDs | Map extension model IDs to endpoint model IDs |
deepseek-bridge.visionModel |
auto | Vision helper model ID |
deepseek-bridge.visionPrompt |
built-in | Prompt used for image description |
deepseek-bridge.retrieval.enabled |
true |
Enable workspace retrieval |
deepseek-bridge.debugMode |
minimal |
Diagnostic verbosity |
deepseek-bridge.allowWorkspaceBaseUrl |
false |
Allow workspace endpoint override |
deepseek-bridge.allowSettingsApiKeyFallback |
false |
Allow API key fallback from settings |
deepseek-bridge.allowSensitiveRequestDumps |
false |
Allow sensitive payload dumps when verbose mode is enabled |
deepseek-bridge.usage.enabled |
true |
Enable local usage and cost display |
deepseek-bridge.usage.showStatusBar |
true |
Show live usage status bar item |
deepseek-bridge.usage.persistSessions |
true |
Save local usage summaries |
deepseek-bridge.usage.maxStoredSessions |
50 |
Max stored usage sessions |
deepseek-bridge.usage.pricingOverrides |
{} |
Override USD-per-million pricing map |