Skip to content

Add support for Deepseek v4 flash #88

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
simonpcouch merged 5 commits into from
Aug 26, 2026
Merged

Add support for Deepseek v4 flash #88

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
dca2432
add DeepSeek V4 Flash model profile
simonpcouch Aug 26, 2026
8226708
simplify DeepSeek thinking support
simonpcouch Aug 26, 2026
78cb790
isolate DeepSeek thinking behavior
simonpcouch Aug 26, 2026
a4ae376
raise DeepSeek context limit
simonpcouch Aug 26, 2026
a04bbd2
remove redundant capability test
simonpcouch Aug 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions packages/ai-config/src/model-capabilities/model-apis-helpers.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,16 @@ const MODEL_APIS_CAPABILITIES: Record<string, Partial<ModelInfo>> = {
maxContextLength: 250_000,
maxInputTokens: 250_000,
},
"deepseek-ai/DeepSeek-V4-Flash-0731": {
family: "deepseek-v4",
thinkingEffortLevels: ["off", "low", "high", "max"],
supportsImages: false,
supportsToolResultImages: false,
supportedInputMediaTypes: [],
maxOutputTokens: 384_000,
maxContextLength: 250_000,
maxInputTokens: 250_000,
Comment on lines +55 to +56

@simonpcouch simonpcouch Aug 26, 2026

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is smaller than the 'real' context length, but artificially bringing this down to help us stay under rate limits.

},
};

/**
Expand Down
14 changes: 14 additions & 0 deletions packages/ai-provider-bridge/src/__tests__/utils.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,7 @@ import {
buildSnowflakeCortexUrl,
buildSnowflakeCortexUrlFromHost,
normalizeProviderBaseUrl,
positAiThinkingRequestFields,
thinkingRequestFields,
} from "../utils";

Expand All @@ -33,6 +34,19 @@ describe("thinkingRequestFields", () => {
});
});

describe("positAiThinkingRequestFields", () => {
it("maps DeepSeek off to none and reuses named efforts", () => {
expect(
positAiThinkingRequestFields("deepseek-ai/DeepSeek-V4-Flash-0731", "off", false),
).toEqual({
reasoning_effort: "none",
});
expect(
positAiThinkingRequestFields("deepseek-ai/DeepSeek-V4-Flash-0731", "high", false),
).toEqual({ reasoning_effort: "high" });
});
});

describe("normalizeProviderBaseUrl", () => {
const HOST = "https://api.anthropic.com";

Expand Down
5 changes: 3 additions & 2 deletions packages/ai-provider-bridge/src/model-clients/PositAiClient.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ import {
isClaudeModel,
isThinkingEnabled,
joinPath,
thinkingRequestFields,
positAiThinkingRequestFields,
} from "../utils";
import {
convertAiSdkStreamToPlatform,
Expand Down Expand Up @@ -238,7 +238,8 @@ export class PositAiClient implements ModelClient {

return convertAiSdkStreamToPlatform(result.fullStream, cleanup);
} else if (normalizedProtocol === "openai-chat") {
const thinkingFields = thinkingRequestFields(
const thinkingFields = positAiThinkingRequestFields(
params.model,
params.thinkingEffort,
params.requiresChatTemplateKwargs ?? false,
);
Expand Down
11 changes: 11 additions & 0 deletions packages/ai-provider-bridge/src/utils.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,6 +41,17 @@ export function thinkingRequestFields(
: undefined;
}

export function positAiThinkingRequestFields(
modelId: string,
effort: string | undefined,
requiresChatTemplateKwargs: boolean,
): Record<string, unknown> | undefined {
if (modelId === "deepseek-ai/DeepSeek-V4-Flash-0731" && effort === "off") {
return { reasoning_effort: "none" };
}
return thinkingRequestFields(effort, requiresChatTemplateKwargs);
}

// ---------------------------------------------------------------------------
// Model ID helpers
// ---------------------------------------------------------------------------
Expand Down