feat(deepseek): support v4 models

Add DeepSeek V4 Flash/Pro support, update Pi packages to 0.70.2, and handle disabled thinking/None by stripping replayed reasoning content.
2026-05-01 06:36:23 +08:00 · 2026-04-24 23:09:36 +08:00
parent 4f4288e3b5
commit 7d1891e6e6
27 changed files with 429 additions and 129 deletions
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -21,7 +21,7 @@ Docs: https://docs.openclaw.ai
 - Agents/subagents: add optional forked context for native `sessions_spawn` runs so agents can let a child inherit the requester transcript when needed, while keeping clean isolated sessions as the default; includes prompt guidance, context-engine hook metadata, docs, and QA coverage.
 - Codex harness: add structured debug logging for embedded harness selection decisions so `/status` stays simple while gateway logs explain auto-selection and Pi fallback reasons. (#70760) Thanks @100yenadmin.
 - Plugin SDK/Codex harness: add provider-owned transport/auth/follow-up seams and harness result classification so Codex-style runtimes can participate in fallback policy without core special-casing. (#70772) Thanks @100yenadmin.
- Dependencies/Pi: update bundled Pi packages to `0.70.0`, use Pi's upstream `gpt-5.5` catalog metadata for OpenAI and OpenAI Codex, and keep only local `gpt-5.5-pro` forward-compat handling.
+- Dependencies/Pi: update bundled Pi packages to `0.70.2`, use Pi's upstream `gpt-5.5` and DeepSeek V4 catalog metadata, and keep only local `gpt-5.5-pro` forward-compat handling.
 - Models/CLI: speed up `openclaw models list --all --provider <id>` for bundled providers with safe static catalogs while keeping live and third-party providers on registry discovery. (#70632) Thanks @shakkernerd.
 - Models/CLI: avoid broad registry enumeration for default `openclaw models list`, reducing default listing latency while preserving configured-row output. (#70883) Thanks @shakkernerd.
 - Models/CLI: split `openclaw models list` row-source orchestration and registry loading into narrower helpers without changing list output behavior. (#70867) Thanks @shakkernerd.
@@ -35,6 +35,7 @@ Docs: https://docs.openclaw.ai
 - Providers/OpenRouter: add image generation and reference-image editing through `image_generate`, so OpenRouter image models work with `OPENROUTER_API_KEY`. Fixes #55066 via #67668. Thanks @notamicrodose.
 - Image generation: let agents request provider-supported quality and output format hints, and pass OpenAI-specific background, moderation, compression, and user hints through the `image_generate` tool. (#70503) Thanks @ottodeng.
 - Plugins/Google Meet: let realtime Meet sessions consult the full OpenClaw agent for deeper answers while staying in the live voice loop.
+- Providers/DeepSeek: add DeepSeek V4 Flash and V4 Pro to the bundled catalog and make V4 Flash the onboarding default.

 ### Fixes

@@ -42,6 +43,7 @@ Docs: https://docs.openclaw.ai
 - Browser/tool: tell agents not to pass per-call `timeoutMs` on existing-session type, evaluate, and other Chrome MCP actions that reject timeout overrides.
 - Codex/GPT-5.4: harden fallback, auth-profile, tool-schema, and replay edge cases across native and embedded runtime paths. (#70743) Thanks @100yenadmin.
 - Voice-call/Telnyx: preserve inbound/outbound callback metadata and read transcription text from Telnyx's current `transcription_data` payload.
+- Providers/DeepSeek: wire V4 thinking controls and OpenAI-compatible replay policy so follow-up turns preserve DeepSeek `reasoning_content`, while the None/off thinking path strips replayed reasoning fields. Fixes #70931.
 - Codex harness: send verbose tool progress to chat channels for native app-server runs, matching the Pi harness `/verbose on` and `/verbose full` behavior. (#70966) Thanks @jalehman.
 - Codex models: fetch paginated Codex app-server model catalogs, mark truncated `/codex models` output, and keep ChatGPT OAuth defaults on the `openai-codex/gpt-5.5` route instead of the OpenAI API-key route.
 - Codex harness: route native `request_user_input` prompts back to the originating chat, preserve queued follow-up answers, and honor newer app-server command approval amendment decisions.
--- a/docs/concepts/model-providers.md
+++ b/docs/concepts/model-providers.md
@@ -235,6 +235,7 @@ See [/providers/kilocode](/providers/kilocode) for setup details.
 | BytePlus                | `byteplus` / `byteplus-plan`     | `BYTEPLUS_API_KEY`                                           | `byteplus-plan/ark-code-latest`                 |
 | Cerebras                | `cerebras`                       | `CEREBRAS_API_KEY`                                           | `cerebras/zai-glm-4.7`                          |
 | Cloudflare AI Gateway   | `cloudflare-ai-gateway`          | `CLOUDFLARE_AI_GATEWAY_API_KEY`                              | —                                               |
+| DeepSeek                | `deepseek`                       | `DEEPSEEK_API_KEY`                                           | `deepseek/deepseek-v4-flash`                    |
 | GitHub Copilot          | `github-copilot`                 | `COPILOT_GITHUB_TOKEN` / `GH_TOKEN` / `GITHUB_TOKEN`         | —                                               |
 | Groq                    | `groq`                           | `GROQ_API_KEY`                                               | —                                               |
 | Hugging Face Inference  | `huggingface`                    | `HUGGINGFACE_HUB_TOKEN` or `HF_TOKEN`                        | `huggingface/deepseek-ai/DeepSeek-R1`           |
--- a/docs/pi.md
+++ b/docs/pi.md
@@ -23,10 +23,10 @@ OpenClaw uses the pi SDK to embed an AI coding agent into its messaging gateway

 ```json
 {
-  "@mariozechner/pi-agent-core": "0.68.1",
-  "@mariozechner/pi-ai": "0.68.1",
-  "@mariozechner/pi-coding-agent": "0.68.1",
-  "@mariozechner/pi-tui": "0.68.1"
+  "@mariozechner/pi-agent-core": "0.70.2",
+  "@mariozechner/pi-ai": "0.70.2",
+  "@mariozechner/pi-coding-agent": "0.70.2",
+  "@mariozechner/pi-tui": "0.70.2"
 }
 ```

--- a/docs/providers/deepseek.md
+++ b/docs/providers/deepseek.md
@@ -26,7 +26,7 @@ read_when:
    openclaw onboard --auth-choice deepseek-api-key
    ```

-    This will prompt for your API key and set `deepseek/deepseek-chat` as the default model.
+    This will prompt for your API key and set `deepseek/deepseek-v4-flash` as the default model.

  </Step>
  <Step title="Verify models are available">
@@ -60,13 +60,17 @@ is available to that process (for example, in `~/.openclaw/.env` or via

 ## Built-in catalog

-| Model ref                    | Name              | Input | Context | Max output | Notes                                             |
-| ---------------------------- | ----------------- | ----- | ------- | ---------- | ------------------------------------------------- |
-| `deepseek/deepseek-chat`     | DeepSeek Chat     | text  | 131,072 | 8,192      | Default model; DeepSeek V3.2 non-thinking surface |
-| `deepseek/deepseek-reasoner` | DeepSeek Reasoner | text  | 131,072 | 65,536     | Reasoning-enabled V3.2 surface                    |
+| Model ref                    | Name              | Input | Context   | Max output | Notes                                      |
+| ---------------------------- | ----------------- | ----- | --------- | ---------- | ------------------------------------------ |
+| `deepseek/deepseek-v4-flash` | DeepSeek V4 Flash | text  | 1,000,000 | 384,000    | Default model; V4 thinking-capable surface |
+| `deepseek/deepseek-v4-pro`   | DeepSeek V4 Pro   | text  | 1,000,000 | 384,000    | V4 thinking-capable surface                |
+| `deepseek/deepseek-chat`     | DeepSeek Chat     | text  | 131,072   | 8,192      | DeepSeek V3.2 non-thinking surface         |
+| `deepseek/deepseek-reasoner` | DeepSeek Reasoner | text  | 131,072   | 65,536     | Reasoning-enabled V3.2 surface             |

 <Tip>
-Both bundled models currently advertise streaming usage compatibility in source.
+V4 models support DeepSeek's `thinking` control. OpenClaw also replays
+DeepSeek `reasoning_content` on follow-up turns so thinking sessions with tool
+calls can continue.
 </Tip>

 ## Config example
@@ -76,7 +80,7 @@ Both bundled models currently advertise streaming usage compatibility in source.
  env: { DEEPSEEK_API_KEY: "sk-..." },
  agents: {
    defaults: {
-      model: { primary: "deepseek/deepseek-chat" },
+      model: { primary: "deepseek/deepseek-v4-flash" },
    },
  },
 }
--- a/extensions/amazon-bedrock-mantle/package.json
+++ b/extensions/amazon-bedrock-mantle/package.json
@@ -7,7 +7,7 @@
  "dependencies": {
    "@anthropic-ai/sdk": "0.90.0",
    "@aws/bedrock-token-generator": "^1.1.0",
-    "@mariozechner/pi-ai": "0.70.0"
+    "@mariozechner/pi-ai": "0.70.2"
  },
  "devDependencies": {
    "@openclaw/plugin-sdk": "workspace:*"
--- a/extensions/anthropic/package.json
+++ b/extensions/anthropic/package.json
@@ -5,7 +5,7 @@
  "description": "OpenClaw Anthropic provider plugin",
  "type": "module",
  "dependencies": {
-    "@mariozechner/pi-ai": "0.70.0"
+    "@mariozechner/pi-ai": "0.70.2"
  },
  "devDependencies": {
    "@openclaw/plugin-sdk": "workspace:*"
--- a/extensions/codex/package.json
+++ b/extensions/codex/package.json
@@ -4,7 +4,7 @@
  "description": "OpenClaw Codex harness and model provider plugin",
  "type": "module",
  "dependencies": {
-    "@mariozechner/pi-coding-agent": "0.70.0",
+    "@mariozechner/pi-coding-agent": "0.70.2",
    "ajv": "^8.18.0",
    "ws": "^8.20.0",
    "zod": "^4.3.6"
--- a/extensions/deepseek/api.ts
+++ b/extensions/deepseek/api.ts
@@ -4,3 +4,4 @@ export {
  DEEPSEEK_MODEL_CATALOG,
 } from "./models.js";
 export { buildDeepSeekProvider } from "./provider-catalog.js";
+export { createDeepSeekV4ThinkingWrapper } from "./stream.js";
--- a/extensions/deepseek/index.test.ts
+++ b/extensions/deepseek/index.test.ts
@@ -1,8 +1,12 @@
+import type { Context, Model } from "@mariozechner/pi-ai";
+import { createAssistantMessageEventStream } from "@mariozechner/pi-ai";
 import { describe, expect, it } from "vitest";
+import { buildOpenAICompletionsParams } from "../../src/agents/openai-transport-stream.js";
 import { resolveProviderPluginChoice } from "../../src/plugins/provider-auth-choice.runtime.js";
 import { registerSingleProviderPlugin } from "../../test/helpers/plugins/plugin-registration.js";
 import { runSingleProviderCatalog } from "../test-support/provider-model-test-helpers.js";
 import deepseekPlugin from "./index.js";
+import { createDeepSeekV4ThinkingWrapper } from "./stream.js";

 describe("deepseek provider plugin", () => {
  it("registers DeepSeek with api-key auth wizard metadata", async () => {
@@ -28,14 +32,181 @@ describe("deepseek provider plugin", () => {
    expect(catalogProvider.api).toBe("openai-completions");
    expect(catalogProvider.baseUrl).toBe("https://api.deepseek.com");
    expect(catalogProvider.models?.map((model) => model.id)).toEqual([
+      "deepseek-v4-flash",
+      "deepseek-v4-pro",
      "deepseek-chat",
      "deepseek-reasoner",
    ]);
+    expect(catalogProvider.models?.find((model) => model.id === "deepseek-v4-flash")).toMatchObject(
+      {
+        reasoning: true,
+        contextWindow: 1_000_000,
+        maxTokens: 384_000,
+        compat: expect.objectContaining({
+          supportsReasoningEffort: true,
+          maxTokensField: "max_tokens",
+        }),
+      },
+    );
    expect(
      catalogProvider.models?.find((model) => model.id === "deepseek-reasoner")?.reasoning,
    ).toBe(true);
  });

+  it("owns OpenAI-compatible replay policy", async () => {
+    const provider = await registerSingleProviderPlugin(deepseekPlugin);
+
+    expect(provider.buildReplayPolicy?.({ modelApi: "openai-completions" } as never)).toMatchObject(
+      {
+        sanitizeToolCallIds: true,
+        toolCallIdMode: "strict",
+        validateGeminiTurns: true,
+        validateAnthropicTurns: true,
+      },
+    );
+  });
+
+  it("maps thinking levels to DeepSeek V4 payload controls", async () => {
+    let capturedPayload: Record<string, unknown> | undefined;
+    const baseStreamFn = (
+      _model: Model<"openai-completions">,
+      _context: Context,
+      options?: { onPayload?: (payload: unknown) => unknown },
+    ) => {
+      capturedPayload = {
+        model: "deepseek-v4-pro",
+        reasoning_effort: "high",
+      };
+      options?.onPayload?.(capturedPayload);
+      const stream = createAssistantMessageEventStream();
+      queueMicrotask(() => stream.end());
+      return stream;
+    };
+
+    const wrapThinkingOff = createDeepSeekV4ThinkingWrapper(baseStreamFn as never, "off");
+    expect(wrapThinkingOff).toBeDefined();
+    wrapThinkingOff?.(
+      {
+        provider: "deepseek",
+        id: "deepseek-v4-pro",
+        api: "openai-completions",
+      } as never,
+      { messages: [] } as never,
+      {},
+    );
+
+    expect(capturedPayload).toMatchObject({ thinking: { type: "disabled" } });
+    expect(capturedPayload).not.toHaveProperty("reasoning_effort");
+
+    const wrapThinkingXhigh = createDeepSeekV4ThinkingWrapper(baseStreamFn as never, "xhigh");
+    expect(wrapThinkingXhigh).toBeDefined();
+    wrapThinkingXhigh?.(
+      {
+        provider: "deepseek",
+        id: "deepseek-v4-pro",
+        api: "openai-completions",
+      } as never,
+      { messages: [] } as never,
+      {},
+    );
+
+    expect(capturedPayload).toMatchObject({
+      thinking: { type: "enabled" },
+      reasoning_effort: "max",
+    });
+  });
+
+  it("strips replayed reasoning_content when DeepSeek V4 thinking is disabled", async () => {
+    let capturedPayload: Record<string, unknown> | undefined;
+    const model = {
+      provider: "deepseek",
+      id: "deepseek-v4-flash",
+      name: "DeepSeek V4 Flash",
+      api: "openai-completions",
+      baseUrl: "https://api.deepseek.com",
+      reasoning: true,
+      input: ["text"],
+      cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+      contextWindow: 1_000_000,
+      maxTokens: 384_000,
+      compat: {
+        supportsUsageInStreaming: true,
+        supportsReasoningEffort: true,
+        maxTokensField: "max_tokens",
+      },
+    } as Model<"openai-completions">;
+    const context = {
+      messages: [
+        { role: "user", content: "hi", timestamp: 1 },
+        {
+          role: "assistant",
+          api: "openai-completions",
+          provider: "deepseek",
+          model: "deepseek-v4-flash",
+          content: [
+            {
+              type: "thinking",
+              thinking: "call reasoning",
+              thinkingSignature: "reasoning_content",
+            },
+            { type: "toolCall", id: "call_1", name: "read", arguments: {} },
+          ],
+          usage: {
+            input: 0,
+            output: 0,
+            cacheRead: 0,
+            cacheWrite: 0,
+            totalTokens: 0,
+            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+          },
+          stopReason: "toolUse",
+          timestamp: 2,
+        },
+        {
+          role: "toolResult",
+          toolCallId: "call_1",
+          toolName: "read",
+          content: [{ type: "text", text: "ok" }],
+          isError: false,
+          timestamp: 3,
+        },
+      ],
+      tools: [
+        {
+          name: "read",
+          description: "Read data",
+          parameters: { type: "object", properties: {}, required: [], additionalProperties: false },
+        },
+      ],
+    } as Context;
+    const baseStreamFn = (
+      streamModel: Model<"openai-completions">,
+      streamContext: Context,
+      options?: { onPayload?: (payload: unknown, model: unknown) => unknown },
+    ) => {
+      capturedPayload = buildOpenAICompletionsParams(streamModel, streamContext, {
+        reasoning: "high",
+      } as never);
+      options?.onPayload?.(capturedPayload, streamModel);
+      const stream = createAssistantMessageEventStream();
+      queueMicrotask(() => stream.end());
+      return stream;
+    };
+
+    const wrapThinkingNone = createDeepSeekV4ThinkingWrapper(
+      baseStreamFn as never,
+      "none" as never,
+    );
+    expect(wrapThinkingNone).toBeDefined();
+    wrapThinkingNone?.(model, context, {});
+
+    expect(capturedPayload).toMatchObject({ thinking: { type: "disabled" } });
+    expect(capturedPayload).not.toHaveProperty("reasoning_effort");
+    expect((capturedPayload?.messages as Array<Record<string, unknown>>)[1]).not.toHaveProperty(
+      "reasoning_content",
+    );
+  });
+
  it("publishes configured DeepSeek models through plugin-owned catalog augmentation", async () => {
    const provider = await registerSingleProviderPlugin(deepseekPlugin);

--- a/extensions/deepseek/index.ts
+++ b/extensions/deepseek/index.ts
@@ -1,7 +1,9 @@
 import { readConfiguredProviderCatalogEntries } from "openclaw/plugin-sdk/provider-catalog-shared";
 import { defineSingleProviderPluginEntry } from "openclaw/plugin-sdk/provider-entry";
+import { buildProviderReplayFamilyHooks } from "openclaw/plugin-sdk/provider-model-shared";
 import { applyDeepSeekConfig, DEEPSEEK_DEFAULT_MODEL_REF } from "./onboard.js";
 import { buildDeepSeekProvider } from "./provider-catalog.js";
+import { createDeepSeekV4ThinkingWrapper } from "./stream.js";

 const PROVIDER_ID = "deepseek";

@@ -42,5 +44,7 @@ export default defineSingleProviderPluginEntry({
      }),
    matchesContextOverflowError: ({ errorMessage }) =>
      /\bdeepseek\b.*(?:input.*too long|context.*exceed)/i.test(errorMessage),
+    ...buildProviderReplayFamilyHooks({ family: "openai-compatible" }),
+    wrapStreamFn: (ctx) => createDeepSeekV4ThinkingWrapper(ctx.streamFn, ctx.thinkingLevel),
  },
 });
--- a/extensions/deepseek/models.ts
+++ b/extensions/deepseek/models.ts
@@ -2,7 +2,7 @@ import type { ModelDefinitionConfig } from "openclaw/plugin-sdk/provider-model-s

 export const DEEPSEEK_BASE_URL = "https://api.deepseek.com";

-// DeepSeek V3.2 API pricing (per 1M tokens)
+// DeepSeek API pricing (per 1M tokens)
 // https://api-docs.deepseek.com/quick_start/pricing
 const DEEPSEEK_V3_2_COST = {
  input: 0.28,
@@ -11,7 +11,49 @@ const DEEPSEEK_V3_2_COST = {
  cacheWrite: 0,
 };

+const DEEPSEEK_V4_PRO_COST = {
+  input: 1.74,
+  output: 3.48,
+  cacheRead: 0.145,
+  cacheWrite: 0,
+};
+
+const DEEPSEEK_V4_FLASH_COST = {
+  input: 0.14,
+  output: 0.28,
+  cacheRead: 0.028,
+  cacheWrite: 0,
+};
+
 export const DEEPSEEK_MODEL_CATALOG: ModelDefinitionConfig[] = [
+  {
+    id: "deepseek-v4-flash",
+    name: "DeepSeek V4 Flash",
+    reasoning: true,
+    input: ["text"],
+    contextWindow: 1_000_000,
+    maxTokens: 384_000,
+    cost: DEEPSEEK_V4_FLASH_COST,
+    compat: {
+      supportsUsageInStreaming: true,
+      supportsReasoningEffort: true,
+      maxTokensField: "max_tokens",
+    },
+  },
+  {
+    id: "deepseek-v4-pro",
+    name: "DeepSeek V4 Pro",
+    reasoning: true,
+    input: ["text"],
+    contextWindow: 1_000_000,
+    maxTokens: 384_000,
+    cost: DEEPSEEK_V4_PRO_COST,
+    compat: {
+      supportsUsageInStreaming: true,
+      supportsReasoningEffort: true,
+      maxTokensField: "max_tokens",
+    },
+  },
  {
    id: "deepseek-chat",
    name: "DeepSeek Chat",
@@ -20,7 +62,7 @@ export const DEEPSEEK_MODEL_CATALOG: ModelDefinitionConfig[] = [
    contextWindow: 131072,
    maxTokens: 8192,
    cost: DEEPSEEK_V3_2_COST,
-    compat: { supportsUsageInStreaming: true },
+    compat: { supportsUsageInStreaming: true, maxTokensField: "max_tokens" },
  },
  {
    id: "deepseek-reasoner",
@@ -30,7 +72,11 @@ export const DEEPSEEK_MODEL_CATALOG: ModelDefinitionConfig[] = [
    contextWindow: 131072,
    maxTokens: 65536,
    cost: DEEPSEEK_V3_2_COST,
-    compat: { supportsUsageInStreaming: true },
+    compat: {
+      supportsUsageInStreaming: true,
+      supportsReasoningEffort: false,
+      maxTokensField: "max_tokens",
+    },
  },
 ];

--- a/extensions/deepseek/onboard.ts
+++ b/extensions/deepseek/onboard.ts
@@ -5,7 +5,7 @@ import {
 } from "openclaw/plugin-sdk/provider-onboard";
 import { buildDeepSeekModelDefinition, DEEPSEEK_BASE_URL, DEEPSEEK_MODEL_CATALOG } from "./api.js";

-export const DEEPSEEK_DEFAULT_MODEL_REF = "deepseek/deepseek-chat";
+export const DEEPSEEK_DEFAULT_MODEL_REF = "deepseek/deepseek-v4-flash";

 export function applyDeepSeekProviderConfig(cfg: OpenClawConfig): OpenClawConfig {
  const models = { ...cfg.agents?.defaults?.models };
--- a/extensions/deepseek/stream.ts
+++ b/extensions/deepseek/stream.ts
@@ -0,0 +1,57 @@
+import type { ProviderWrapStreamFnContext } from "openclaw/plugin-sdk/plugin-entry";
+import { streamWithPayloadPatch } from "openclaw/plugin-sdk/provider-stream-shared";
+
+type DeepSeekThinkingLevel = ProviderWrapStreamFnContext["thinkingLevel"];
+
+function isDeepSeekV4ModelId(modelId: unknown): boolean {
+  return modelId === "deepseek-v4-flash" || modelId === "deepseek-v4-pro";
+}
+
+function isDisabledThinkingLevel(thinkingLevel: DeepSeekThinkingLevel): boolean {
+  const normalized = typeof thinkingLevel === "string" ? thinkingLevel.toLowerCase() : "";
+  return normalized === "off" || normalized === "none";
+}
+
+function resolveDeepSeekReasoningEffort(thinkingLevel: DeepSeekThinkingLevel): "high" | "max" {
+  return thinkingLevel === "xhigh" || thinkingLevel === "max" ? "max" : "high";
+}
+
+function stripDeepSeekReasoningContent(payload: Record<string, unknown>): void {
+  if (!Array.isArray(payload.messages)) {
+    return;
+  }
+  for (const message of payload.messages) {
+    if (!message || typeof message !== "object") {
+      continue;
+    }
+    delete (message as Record<string, unknown>).reasoning_content;
+  }
+}
+
+export function createDeepSeekV4ThinkingWrapper(
+  baseStreamFn: ProviderWrapStreamFnContext["streamFn"],
+  thinkingLevel: DeepSeekThinkingLevel,
+): ProviderWrapStreamFnContext["streamFn"] {
+  if (!baseStreamFn) {
+    return undefined;
+  }
+  const underlying = baseStreamFn;
+  return (model, context, options) => {
+    if (model.provider !== "deepseek" || !isDeepSeekV4ModelId(model.id)) {
+      return underlying(model, context, options);
+    }
+
+    return streamWithPayloadPatch(underlying, model, context, options, (payload) => {
+      if (isDisabledThinkingLevel(thinkingLevel)) {
+        payload.thinking = { type: "disabled" };
+        delete payload.reasoning_effort;
+        delete payload.reasoning;
+        stripDeepSeekReasoningContent(payload);
+        return;
+      }
+
+      payload.thinking = { type: "enabled" };
+      payload.reasoning_effort = resolveDeepSeekReasoningEffort(thinkingLevel);
+    });
+  };
+}
--- a/extensions/fireworks/package.json
+++ b/extensions/fireworks/package.json
@@ -5,7 +5,7 @@
  "description": "OpenClaw Fireworks provider plugin",
  "type": "module",
  "dependencies": {
-    "@mariozechner/pi-ai": "0.70.0"
+    "@mariozechner/pi-ai": "0.70.2"
  },
  "devDependencies": {
    "@openclaw/plugin-sdk": "workspace:*"
--- a/extensions/github-copilot/package.json
+++ b/extensions/github-copilot/package.json
@@ -6,7 +6,7 @@
  "type": "module",
  "dependencies": {
    "@clack/prompts": "^1.2.0",
-    "@mariozechner/pi-ai": "0.70.0"
+    "@mariozechner/pi-ai": "0.70.2"
  },
  "devDependencies": {
    "@openclaw/plugin-sdk": "workspace:*"
--- a/extensions/google/package.json
+++ b/extensions/google/package.json
@@ -6,7 +6,7 @@
  "type": "module",
  "dependencies": {
    "@google/genai": "^1.50.1",
-    "@mariozechner/pi-ai": "0.70.0"
+    "@mariozechner/pi-ai": "0.70.2"
  },
  "devDependencies": {
    "@openclaw/plugin-sdk": "workspace:*"
--- a/extensions/kimi-coding/package.json
+++ b/extensions/kimi-coding/package.json
@@ -5,7 +5,7 @@
  "description": "OpenClaw Kimi provider plugin",
  "type": "module",
  "dependencies": {
-    "@mariozechner/pi-ai": "0.70.0"
+    "@mariozechner/pi-ai": "0.70.2"
  },
  "devDependencies": {
    "@openclaw/plugin-sdk": "workspace:*"
--- a/extensions/lmstudio/package.json
+++ b/extensions/lmstudio/package.json
@@ -5,7 +5,7 @@
  "description": "OpenClaw LM Studio provider plugin",
  "type": "module",
  "dependencies": {
-    "@mariozechner/pi-ai": "0.70.0"
+    "@mariozechner/pi-ai": "0.70.2"
  },
  "openclaw": {
    "extensions": [
--- a/extensions/ollama/package.json
+++ b/extensions/ollama/package.json
@@ -5,7 +5,7 @@
  "description": "OpenClaw Ollama provider plugin",
  "type": "module",
  "dependencies": {
-    "@mariozechner/pi-ai": "0.70.0",
+    "@mariozechner/pi-ai": "0.70.2",
    "typebox": "1.1.28"
  },
  "devDependencies": {
--- a/extensions/openai/package.json
+++ b/extensions/openai/package.json
@@ -5,7 +5,7 @@
  "description": "OpenClaw OpenAI provider plugins",
  "type": "module",
  "dependencies": {
-    "@mariozechner/pi-ai": "0.70.0",
+    "@mariozechner/pi-ai": "0.70.2",
    "ws": "^8.20.0"
  },
  "devDependencies": {
--- a/extensions/xai/package.json
+++ b/extensions/xai/package.json
@@ -5,7 +5,7 @@
  "description": "OpenClaw xAI plugin",
  "type": "module",
  "dependencies": {
-    "@mariozechner/pi-ai": "0.70.0",
+    "@mariozechner/pi-ai": "0.70.2",
    "typebox": "1.1.28",
    "ws": "^8.20.0"
  },
--- a/package.json
+++ b/package.json
@@ -1582,10 +1582,10 @@
    "@anthropic-ai/vertex-sdk": "^0.16.0",
    "@clack/prompts": "^1.2.0",
    "@lydell/node-pty": "1.2.0-beta.12",
-    "@mariozechner/pi-agent-core": "0.70.0",
-    "@mariozechner/pi-ai": "0.70.0",
-    "@mariozechner/pi-coding-agent": "0.70.0",
-    "@mariozechner/pi-tui": "0.70.0",
+    "@mariozechner/pi-agent-core": "0.70.2",
+    "@mariozechner/pi-ai": "0.70.2",
+    "@mariozechner/pi-coding-agent": "0.70.2",
+    "@mariozechner/pi-tui": "0.70.2",
    "@modelcontextprotocol/sdk": "1.29.0",
    "@mozilla/readability": "^0.6.0",
    "@vincentkoc/qrcode-tui": "0.2.1",
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@@ -52,17 +52,17 @@ importers:
        specifier: 1.2.0-beta.12
        version: 1.2.0-beta.12
      '@mariozechner/pi-agent-core':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
      '@mariozechner/pi-coding-agent':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
      '@mariozechner/pi-tui':
-        specifier: 0.70.0
-        version: 0.70.0
+        specifier: 0.70.2
+        version: 0.70.2
      '@modelcontextprotocol/sdk':
        specifier: 1.29.0
        version: 1.29.0(zod@4.3.6)
@@ -277,8 +277,8 @@ importers:
        specifier: ^1.1.0
        version: 1.1.0
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
    devDependencies:
      '@openclaw/plugin-sdk':
        specifier: workspace:*
@@ -287,8 +287,8 @@ importers:
  extensions/anthropic:
    dependencies:
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
    devDependencies:
      '@openclaw/plugin-sdk':
        specifier: workspace:*
@@ -384,8 +384,8 @@ importers:
  extensions/codex:
    dependencies:
      '@mariozechner/pi-coding-agent':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
      ajv:
        specifier: ^8.18.0
        version: 8.18.0
@@ -578,8 +578,8 @@ importers:
  extensions/fireworks:
    dependencies:
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
    devDependencies:
      '@openclaw/plugin-sdk':
        specifier: workspace:*
@@ -591,8 +591,8 @@ importers:
        specifier: ^1.2.0
        version: 1.2.0
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
    devDependencies:
      '@openclaw/plugin-sdk':
        specifier: workspace:*
@@ -604,8 +604,8 @@ importers:
        specifier: ^1.50.1
        version: 1.50.1(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
    devDependencies:
      '@openclaw/plugin-sdk':
        specifier: workspace:*
@@ -685,8 +685,8 @@ importers:
  extensions/kimi-coding:
    dependencies:
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
    devDependencies:
      '@openclaw/plugin-sdk':
        specifier: workspace:*
@@ -727,8 +727,8 @@ importers:
  extensions/lmstudio:
    dependencies:
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)

  extensions/lobster:
    dependencies:
@@ -954,8 +954,8 @@ importers:
  extensions/ollama:
    dependencies:
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
      typebox:
        specifier: 1.1.28
        version: 1.1.28
@@ -973,8 +973,8 @@ importers:
  extensions/openai:
    dependencies:
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
      ws:
        specifier: ^8.20.0
        version: 8.20.0
@@ -1380,8 +1380,8 @@ importers:
  extensions/xai:
    dependencies:
      '@mariozechner/pi-ai':
-        specifier: 0.70.0
-        version: 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+        specifier: 0.70.2
+        version: 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
      typebox:
        specifier: 1.1.28
        version: 1.1.28
@@ -2693,94 +2693,94 @@ packages:
  '@lydell/node-pty@1.2.0-beta.12':
    resolution: {integrity: sha512-qIK890UwPupoj07osVvgOIa++1mxeHbcGry4PKRHhNVNs81V2SCG34eJr46GybiOmBtc8Sj5PB1/GGM5PL549g==}

-  '@mariozechner/clipboard-darwin-arm64@0.3.2':
-    resolution: {integrity: sha512-uBf6K7Je1ihsgvmWxA8UCGCeI+nbRVRXoarZdLjl6slz94Zs1tNKFZqx7aCI5O1i3e0B6ja82zZ06BWrl0MCVw==}
+  '@mariozechner/clipboard-darwin-arm64@0.3.3':
+    resolution: {integrity: sha512-+zhuZGXqVrdkbIRdnwiZNbTJ7V3elq/A+C5d5laJoyhJgWs41eO5NUMkBkj6f23F2L4PRXEhdn5/ktlPx+bG3Q==}
    engines: {node: '>= 10'}
    cpu: [arm64]
    os: [darwin]

-  '@mariozechner/clipboard-darwin-universal@0.3.2':
-    resolution: {integrity: sha512-mxSheKTW2U9LsBdXy0SdmdCAE5HqNS9QUmpNHLnfJ+SsbFKALjEZc5oRrVMXxGQSirDvYf5bjmRyT0QYYonnlg==}
+  '@mariozechner/clipboard-darwin-universal@0.3.3':
+    resolution: {integrity: sha512-x9aRfTyndVqpEQ44LNNCK/EXZd9y8rWkLQgNhmWpby9PXrjPhNxfjUc2Db4mt4nJjU/4zzO8F5v/XyzlUGSdhQ==}
    engines: {node: '>= 10'}
    os: [darwin]

-  '@mariozechner/clipboard-darwin-x64@0.3.2':
-    resolution: {integrity: sha512-U1BcVEoidvwIp95+HJswSW+xr28EQiHR7rZjH6pn8Sja5yO4Yoe3yCN0Zm8Lo72BbSOK/fTSq0je7CJpaPCspg==}
+  '@mariozechner/clipboard-darwin-x64@0.3.3':
+    resolution: {integrity: sha512-6ut/NawB0KiYPCwrirgNp6Br62LntL978q7G6d/Rs2pmPvQb53bP96eUMYl+Y3a7Qk13bGZ4w9rVPFxRE9m9ag==}
    engines: {node: '>= 10'}
    cpu: [x64]
    os: [darwin]

-  '@mariozechner/clipboard-linux-arm64-gnu@0.3.2':
-    resolution: {integrity: sha512-BsinwG3yWTIjdgNCxsFlip7LkfwPk+ruw/aFCXHUg/fb5XC/Ksp+YMQ7u0LUtiKzIv/7LMXgZInJQH6gxbAaqQ==}
+  '@mariozechner/clipboard-linux-arm64-gnu@0.3.3':
+    resolution: {integrity: sha512-gf3dH4kBddU1AOyHVB53mjLUFfJAKlTmxTMw51jdeg7eE7IjfEBXVvM4bifMtBxbWkT0eA0FUZ1C0KQ6Z5l6pw==}
    engines: {node: '>= 10'}
    cpu: [arm64]
    os: [linux]
    libc: [glibc]

-  '@mariozechner/clipboard-linux-arm64-musl@0.3.2':
-    resolution: {integrity: sha512-0/Gi5Xq2V6goXBop19ePoHvXsmJD9SzFlO3S+d6+T2b+BlPcpOu3Oa0wTjl+cZrLAAEzA86aPNBI+VVAFDFPKw==}
+  '@mariozechner/clipboard-linux-arm64-musl@0.3.3':
+    resolution: {integrity: sha512-o1paj2+zmAQ/LaPS85XJCxhNowNQpxYM2cGY6pWvB5Kqmz6hZjl6CzDg5tbf1hZkn/Em6jpOaE2UtMxKdELBDA==}
    engines: {node: '>= 10'}
    cpu: [arm64]
    os: [linux]
    libc: [musl]

-  '@mariozechner/clipboard-linux-riscv64-gnu@0.3.2':
-    resolution: {integrity: sha512-2AFFiXB24qf0zOZsxI1GJGb9wQGlOJyN6UwoXqmKS3dpQi/l6ix30IzDDA4c4ZcCcx4D+9HLYXhC1w7Sov8pXA==}
+  '@mariozechner/clipboard-linux-riscv64-gnu@0.3.3':
+    resolution: {integrity: sha512-dkEhE4ekePJwMbBq9HP1//CFMNmDzA/iV9AXqBfvL5CWmmDIRXqh4A3YZt3tWO/HdMerX+xNCEiR7WiOsIG+UA==}
    engines: {node: '>= 10'}
    cpu: [riscv64]
    os: [linux]
    libc: [glibc]

-  '@mariozechner/clipboard-linux-x64-gnu@0.3.2':
-    resolution: {integrity: sha512-v6fVnsn7WMGg73Dab8QMwyFce7tzGfgEixKgzLP8f1GJqkJZi5zO4k4FOHzSgUufgLil63gnxvMpjWkgfeQN7A==}
+  '@mariozechner/clipboard-linux-x64-gnu@0.3.3':
+    resolution: {integrity: sha512-lT2yANtTLlEtFBIH3uGoRa/CQas/eBoLNi3qr9axQFoRgF4RGPSJ66yHOSnMECBneTIb1Iqv3UxokTfX27CdoQ==}
    engines: {node: '>= 10'}
    cpu: [x64]
    os: [linux]
    libc: [glibc]

-  '@mariozechner/clipboard-linux-x64-musl@0.3.2':
-    resolution: {integrity: sha512-xVUtnoMQ8v2JVyfJLKKXACA6avdnchdbBkTsZs8BgJQo29qwCp5NIHAUO8gbJ40iaEGToW5RlmVk2M9V0HsHEw==}
+  '@mariozechner/clipboard-linux-x64-musl@0.3.3':
+    resolution: {integrity: sha512-saq/MCB0QHK/7ZZLjAZ0QkbY944dyjOsur8gneGCfMitt+GOiE1CU4OUipHC4b6x8UDY9bRLsR4aBaxu22OFPA==}
    engines: {node: '>= 10'}
    cpu: [x64]
    os: [linux]
    libc: [musl]

-  '@mariozechner/clipboard-win32-arm64-msvc@0.3.2':
-    resolution: {integrity: sha512-AEgg95TNi8TGgak2wSXZkXKCvAUTjWoU1Pqb0ON7JHrX78p616XUFNTJohtIon3e0w6k0pYPZeCuqRCza/Tqeg==}
+  '@mariozechner/clipboard-win32-arm64-msvc@0.3.3':
+    resolution: {integrity: sha512-cGuvSj0/2X2w983yEcKw+i+r1EBej6ZZIN+fXG3eY2G/HaIQpbXpLvMxKyZ9LKtbZx+Z6q/gELEoSBMLML6BaQ==}
    engines: {node: '>= 10'}
    cpu: [arm64]
    os: [win32]

-  '@mariozechner/clipboard-win32-x64-msvc@0.3.2':
-    resolution: {integrity: sha512-tGRuYpZwDOD7HBrCpyRuhGnHHSCknELvqwKKUG4JSfSB7JIU7LKRh6zx6fMUOQd8uISK35TjFg5UcNih+vJhFA==}
+  '@mariozechner/clipboard-win32-x64-msvc@0.3.3':
+    resolution: {integrity: sha512-5hvaEq/bgYovTIGx43O/S7loIHYV3ue90WcV1dz0wdMXroVKZKeU/yfwM0PALQA1OcrEHiGXGySFReXr72lGtA==}
    engines: {node: '>= 10'}
    cpu: [x64]
    os: [win32]

-  '@mariozechner/clipboard@0.3.2':
-    resolution: {integrity: sha512-IHQpksNjo7EAtGuHFU+tbWDp5LarH3HU/8WiB9O70ZEoBPHOg0/6afwSLK0QyNMMmx4Bpi/zl6+DcBXe95nWYA==}
+  '@mariozechner/clipboard@0.3.3':
+    resolution: {integrity: sha512-e7jASirzfm+ROiOGFh843+cFZTy3DfzP+jldCvh8RnEk0C3QihDTn7dd7Yh7KAJydwIJ18FJSZ2swHvCJhk18g==}
    engines: {node: '>= 10'}

  '@mariozechner/jiti@2.6.5':
    resolution: {integrity: sha512-faGUlTcXka5l7rv0lP3K3vGW/ejRuOS24RR2aSFWREUQqzjgdsuWNo/IiPqL3kWRGt6Ahl2+qcDAwtdeWeuGUw==}
    hasBin: true

-  '@mariozechner/pi-agent-core@0.70.0':
-    resolution: {integrity: sha512-ZwfM5QPvSwza/apZhPIXjrI/blJBFqbVpK30ma4zNwH8VAyseKlzGDExCx/k+81Xydg60sQuG2BQVkYGmofuSg==}
+  '@mariozechner/pi-agent-core@0.70.2':
+    resolution: {integrity: sha512-g1hIdKyDwmQOoBGO0R4OhpemKeMENeK0vE5FJtuQKqEcsdCAkVBgZAK6aZUARYZVxMA718JS6WPLFWoddzjD7g==}
    engines: {node: '>=20.0.0'}

-  '@mariozechner/pi-ai@0.70.0':
-    resolution: {integrity: sha512-lVT9bb0eFkNr5YXvZ5r00TNA5r110fOO8uJV9VLCQ5GdtunWIjcptWitzIjjl2MF0/NDs7Kb2EwZctXQWWP7eA==}
+  '@mariozechner/pi-ai@0.70.2':
+    resolution: {integrity: sha512-+30LRPjXsXF+oI96DvGWMbdPGeqoLJvadh6UPev7wx2DzhC9FEqXkQcoMZ0usbCm7E9pl8ua8a9s/pQ5ikaUbg==}
    engines: {node: '>=20.0.0'}
    hasBin: true

-  '@mariozechner/pi-coding-agent@0.70.0':
-    resolution: {integrity: sha512-Sw5odG9BYIcRItb/o4Gmq0nSIgoWfx61Isjk3Gk4KqocxHZAOwZZYQ4mgb4GCsevqOMmAzX/H6PC52/TiN76fw==}
+  '@mariozechner/pi-coding-agent@0.70.2':
+    resolution: {integrity: sha512-asfNqV89HKAmKvJ1wENBY/UQMIf77kLtkzBrvXnMQV4YbH7D/6KT+VeVzPG6zm5PAZP2UtdLY9B9Cge7IxH37w==}
    engines: {node: '>=20.6.0'}
    hasBin: true

-  '@mariozechner/pi-tui@0.70.0':
-    resolution: {integrity: sha512-x/CwIMP8v9KNrmgEFA0+AWIwSWeNAitEI4eVQtQ6q2a0PpE+vx1+j2oc+iDPe7E1YqrMHXaNlHJVCaVAv/UYrg==}
+  '@mariozechner/pi-tui@0.70.2':
+    resolution: {integrity: sha512-PtKC0NepnrYcqMx6MXkWTrBzC9tI62KeC6w940oT46lCbfvgmfqXciR15+9BZpxxc1H4jd3CMrKsmOPVeUqZ0A==}
    engines: {node: '>=20.0.0'}

  '@matrix-org/matrix-sdk-crypto-nodejs@0.4.0':
@@ -8705,17 +8705,17 @@ snapshots:
    dependencies:
      '@aws-crypto/sha256-browser': 5.2.0
      '@aws-crypto/sha256-js': 5.2.0
-      '@aws-sdk/core': 3.974.3
+      '@aws-sdk/core': 3.974.2
      '@aws-sdk/middleware-host-header': 3.972.10
      '@aws-sdk/middleware-logger': 3.972.10
      '@aws-sdk/middleware-recursion-detection': 3.972.11
-      '@aws-sdk/middleware-user-agent': 3.972.33
-      '@aws-sdk/region-config-resolver': 3.972.13
+      '@aws-sdk/middleware-user-agent': 3.972.32
+      '@aws-sdk/region-config-resolver': 3.972.12
      '@aws-sdk/signature-v4-multi-region': 3.996.19
      '@aws-sdk/types': 3.973.8
-      '@aws-sdk/util-endpoints': 3.996.8
+      '@aws-sdk/util-endpoints': 3.996.7
      '@aws-sdk/util-user-agent-browser': 3.972.10
-      '@aws-sdk/util-user-agent-node': 3.973.19
+      '@aws-sdk/util-user-agent-node': 3.973.18
      '@smithy/config-resolver': 4.4.17
      '@smithy/core': 3.23.16
      '@smithy/fetch-http-handler': 5.3.17
@@ -9909,48 +9909,48 @@ snapshots:
      '@lydell/node-pty-win32-arm64': 1.2.0-beta.12
      '@lydell/node-pty-win32-x64': 1.2.0-beta.12

-  '@mariozechner/clipboard-darwin-arm64@0.3.2':
+  '@mariozechner/clipboard-darwin-arm64@0.3.3':
    optional: true

-  '@mariozechner/clipboard-darwin-universal@0.3.2':
+  '@mariozechner/clipboard-darwin-universal@0.3.3':
    optional: true

-  '@mariozechner/clipboard-darwin-x64@0.3.2':
+  '@mariozechner/clipboard-darwin-x64@0.3.3':
    optional: true

-  '@mariozechner/clipboard-linux-arm64-gnu@0.3.2':
+  '@mariozechner/clipboard-linux-arm64-gnu@0.3.3':
    optional: true

-  '@mariozechner/clipboard-linux-arm64-musl@0.3.2':
+  '@mariozechner/clipboard-linux-arm64-musl@0.3.3':
    optional: true

-  '@mariozechner/clipboard-linux-riscv64-gnu@0.3.2':
+  '@mariozechner/clipboard-linux-riscv64-gnu@0.3.3':
    optional: true

-  '@mariozechner/clipboard-linux-x64-gnu@0.3.2':
+  '@mariozechner/clipboard-linux-x64-gnu@0.3.3':
    optional: true

-  '@mariozechner/clipboard-linux-x64-musl@0.3.2':
+  '@mariozechner/clipboard-linux-x64-musl@0.3.3':
    optional: true

-  '@mariozechner/clipboard-win32-arm64-msvc@0.3.2':
+  '@mariozechner/clipboard-win32-arm64-msvc@0.3.3':
    optional: true

-  '@mariozechner/clipboard-win32-x64-msvc@0.3.2':
+  '@mariozechner/clipboard-win32-x64-msvc@0.3.3':
    optional: true

-  '@mariozechner/clipboard@0.3.2':
+  '@mariozechner/clipboard@0.3.3':
    optionalDependencies:
-      '@mariozechner/clipboard-darwin-arm64': 0.3.2
-      '@mariozechner/clipboard-darwin-universal': 0.3.2
-      '@mariozechner/clipboard-darwin-x64': 0.3.2
-      '@mariozechner/clipboard-linux-arm64-gnu': 0.3.2
-      '@mariozechner/clipboard-linux-arm64-musl': 0.3.2
-      '@mariozechner/clipboard-linux-riscv64-gnu': 0.3.2
-      '@mariozechner/clipboard-linux-x64-gnu': 0.3.2
-      '@mariozechner/clipboard-linux-x64-musl': 0.3.2
-      '@mariozechner/clipboard-win32-arm64-msvc': 0.3.2
-      '@mariozechner/clipboard-win32-x64-msvc': 0.3.2
+      '@mariozechner/clipboard-darwin-arm64': 0.3.3
+      '@mariozechner/clipboard-darwin-universal': 0.3.3
+      '@mariozechner/clipboard-darwin-x64': 0.3.3
+      '@mariozechner/clipboard-linux-arm64-gnu': 0.3.3
+      '@mariozechner/clipboard-linux-arm64-musl': 0.3.3
+      '@mariozechner/clipboard-linux-riscv64-gnu': 0.3.3
+      '@mariozechner/clipboard-linux-x64-gnu': 0.3.3
+      '@mariozechner/clipboard-linux-x64-musl': 0.3.3
+      '@mariozechner/clipboard-win32-arm64-msvc': 0.3.3
+      '@mariozechner/clipboard-win32-x64-msvc': 0.3.3
    optional: true

  '@mariozechner/jiti@2.6.5':
@@ -9958,9 +9958,9 @@ snapshots:
      std-env: 3.10.0
      yoctocolors: 2.1.2

-  '@mariozechner/pi-agent-core@0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)':
+  '@mariozechner/pi-agent-core@0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)':
    dependencies:
-      '@mariozechner/pi-ai': 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+      '@mariozechner/pi-ai': 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
      typebox: 1.1.28
    transitivePeerDependencies:
      - '@modelcontextprotocol/sdk'
@@ -9971,7 +9971,7 @@ snapshots:
      - ws
      - zod

-  '@mariozechner/pi-ai@0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)':
+  '@mariozechner/pi-ai@0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)':
    dependencies:
      '@anthropic-ai/sdk': 0.90.0(zod@4.3.6)
      '@aws-sdk/client-bedrock-runtime': 3.1034.0
@@ -9993,12 +9993,12 @@ snapshots:
      - ws
      - zod

-  '@mariozechner/pi-coding-agent@0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)':
+  '@mariozechner/pi-coding-agent@0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)':
    dependencies:
      '@mariozechner/jiti': 2.6.5
-      '@mariozechner/pi-agent-core': 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
-      '@mariozechner/pi-ai': 0.70.0(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
-      '@mariozechner/pi-tui': 0.70.0
+      '@mariozechner/pi-agent-core': 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+      '@mariozechner/pi-ai': 0.70.2(@modelcontextprotocol/sdk@1.29.0(zod@4.3.6))(ws@8.20.0)(zod@4.3.6)
+      '@mariozechner/pi-tui': 0.70.2
      '@silvia-odwyer/photon-node': 0.3.4
      chalk: 5.6.2
      cli-highlight: 2.1.11
@@ -10017,7 +10017,7 @@ snapshots:
      uuid: 14.0.0
      yaml: 2.8.3
    optionalDependencies:
-      '@mariozechner/clipboard': 0.3.2
+      '@mariozechner/clipboard': 0.3.3
    transitivePeerDependencies:
      - '@modelcontextprotocol/sdk'
      - aws-crt
@@ -10027,7 +10027,7 @@ snapshots:
      - ws
      - zod

-  '@mariozechner/pi-tui@0.70.0':
+  '@mariozechner/pi-tui@0.70.2':
    dependencies:
      '@types/mime-types': 2.1.4
      chalk: 5.6.2
--- a/src/agents/openai-completions-compat.ts
+++ b/src/agents/openai-completions-compat.ts
@@ -16,7 +16,7 @@ export type OpenAICompletionsCompatDefaults = {
  supportsReasoningEffort: boolean;
  supportsUsageInStreaming: boolean;
  maxTokensField: "max_completion_tokens" | "max_tokens";
-  thinkingFormat: "openai" | "openrouter" | "zai";
+  thinkingFormat: "openai" | "openrouter" | "deepseek" | "zai";
  visibleReasoningDetailTypes: string[];
  supportsStrictMode: boolean;
 };
@@ -72,6 +72,9 @@ export function resolveOpenAICompletionsCompatDefaults(
  const isZai =
    endpointClass === "zai-native" ||
    (isDefaultRoute && isDefaultRouteProvider(input.provider, "zai"));
+  const isDeepSeek =
+    endpointClass === "deepseek-native" ||
+    (isDefaultRoute && isDefaultRouteProvider(input.provider, "deepseek"));
  const isNonStandard =
    endpointClass === "cerebras-native" ||
    endpointClass === "chutes-native" ||
@@ -105,7 +108,13 @@ export function resolveOpenAICompletionsCompatDefaults(
      supportsKnownLocalStreamingUsage ||
      (!isNonStandard && (!usesConfiguredNonOpenAIEndpoint || supportsNativeStreamingUsageCompat)),
    maxTokensField: usesMaxTokens ? "max_tokens" : "max_completion_tokens",
-    thinkingFormat: isZai ? "zai" : isOpenRouterLike ? "openrouter" : "openai",
+    thinkingFormat: isDeepSeek
+      ? "deepseek"
+      : isZai
+        ? "zai"
+        : isOpenRouterLike
+          ? "openrouter"
+          : "openai",
    visibleReasoningDetailTypes: isOpenRouterLike ? ["response.output_text", "response.text"] : [],
    supportsStrictMode: !isZai && !usesConfiguredNonOpenAIEndpoint,
  };
--- a/src/config/schema.base.generated.ts
+++ b/src/config/schema.base.generated.ts
@@ -2900,6 +2900,10 @@ export const GENERATED_BASE_CONFIG_SCHEMA: BaseConfigSchemaResponse = {
                                type: "string",
                                const: "openrouter",
                              },
+                              {
+                                type: "string",
+                                const: "deepseek",
+                              },
                              {
                                type: "string",
                                const: "zai",
--- a/src/config/zod-schema.core.ts
+++ b/src/config/zod-schema.core.ts
@@ -202,6 +202,7 @@ export const ModelCompatSchema = z
      .union([
        z.literal("openai"),
        z.literal("openrouter"),
+        z.literal("deepseek"),
        z.literal("zai"),
        z.literal("qwen"),
        z.literal("qwen-chat-template"),
--- a/ui/src/ui/thinking.ts
+++ b/ui/src/ui/thinking.ts
@@ -42,7 +42,7 @@ export function normalizeThinkLevel(raw?: string | null): string | undefined {
  if (collapsed === "xhigh" || collapsed === "extrahigh") {
    return "xhigh";
  }
-  if (key === "off") {
+  if (key === "off" || key === "none") {
    return "off";
  }
  if (["on", "enable", "enabled"].includes(key)) {