modelrig 0.1.0 → 0.3.0

package/dist/adapters/anthropic.js

@@ -5,9 +5,10 @@
  * CALL: the route schema becomes the single tool's input_schema and
  * tool_choice pins it, so the accumulated input_json_delta stream IS the
  * JSON output (Anthropic's recommended structured-output pattern; the run
- * loop still ajv-validates like every provider). Prompt caching is explicit
- * cache_control on Anthropic and is NOT wired yet — all cache capability
- * flags stay false until it is.
+ * loop still ajv-validates like every provider). Prompt caching (WS-C) is
+ * explicit `cache_control` on Anthropic: when the request opts in, the system
+ * prefix is marked with an ephemeral breakpoint (buildAnthropicParams) and the
+ * `context_cache_explicit` capability is advertised per the cache-policy registry.
  */
 var __importDefault = (this && this.__importDefault) || function (mod) {
     return (mod && mod.__esModule) ? mod : { "default": mod };
@@ -17,8 +18,11 @@ exports.isStrictSchemaRejection = isStrictSchemaRejection;
 exports.anthropicSupports = anthropicSupports;
 exports.classifyAnthropicFailure = classifyAnthropicFailure;
 exports.buildAnthropicParams = buildAnthropicParams;
+exports.normalizeAnthropicInputUsage = normalizeAnthropicInputUsage;
 exports.createAnthropicAdapter = createAnthropicAdapter;
 const sdk_1 = __importDefault(require("@anthropic-ai/sdk"));
+const tool_mapping_1 = require("./tool-mapping");
+const sampling_1 = require("./sampling");
 const DEFAULT_MAX_TOKENS = 32000;
 const OUTPUT_TOOL_NAME = "route_output";
 /** Strict tool use (constrained decoding) — public beta. Schemas that
@@ -34,13 +38,14 @@ function anthropicSupports(feature, _model) {
     switch (feature) {
         case "structured_native": // forced tool call with the route schema
         case "json_mode":
+        case "context_cache_explicit": // cache_control ephemeral blocks (WS-C) — cache-policy registry truth
             return true;
         case "grounded_native": // web-search tool exists but is not wired
-        case "context_cache_explicit": // cache_control not wired yet
-        case "prompt_cache_key":
+        case "prompt_cache_key": // anthropic ignores keys — caching is prefix-based via cache_control
         case "prefix_cache_implicit":
         case "tier_flex":
         case "trace_visible": // extended thinking not enabled by this adapter
+        case "zero_retention": // never a static adapter attestation — set only from the credential-scoped registry facts (claims.ts)
             return false;
     }
 }
@@ -73,31 +78,85 @@ function classifyAnthropicFailure(message) {
     return { cls: "network" };
 }
 function buildAnthropicParams(model, req, strict = false) {
+    // Structured output is a FORCED tool call on the route schema (existing
+    // path). Otherwise, caller-supplied `tools` (spec §4.4) map to Anthropic's
+    // tool_use — the two are mutually exclusive here (a structured route does
+    // not also expose free tools in Phase A).
+    const toolBlock = req.outputSchema !== null
+        ? {
+            tools: [
+                {
+                    name: OUTPUT_TOOL_NAME,
+                    description: "Emit the structured output for this route.",
+                    input_schema: req.outputSchema,
+                    ...(strict ? { strict: true } : {}),
+                },
+            ],
+            tool_choice: { type: "tool", name: OUTPUT_TOOL_NAME },
+        }
+        : req.tools !== undefined && req.tools.length > 0
+            ? {
+                tools: (0, tool_mapping_1.toAnthropicTools)(req.tools),
+                ...((0, tool_mapping_1.toAnthropicToolChoice)(req.toolChoice) !== undefined
+                    ? { tool_choice: (0, tool_mapping_1.toAnthropicToolChoice)(req.toolChoice) }
+                    : {}),
+            }
+            : {};
+    // Prompt caching (WS-C): when the request opts in (`cache` set by the run
+    // loop's policy-gated stamp), mark the system prompt — the stable prefix in
+    // ModelRig's system-carries-the-task convention — with a `cache_control`
+    // ephemeral breakpoint. Anthropic caches the prefix up to and including the
+    // marked block (default 5m TTL, refreshes free on read); the `cache.key` is
+    // ignored (caching is prefix-based, not key-based — that is why the telemetry
+    // `cache_key` for anthropic is null). Extended 1h TTL is a documented
+    // follow-up (needs the extended-cache beta header; WS-A already accounts the
+    // 5m/1h split when the provider reports it).
+    const wantsCache = req.cache?.key !== undefined && req.cache.key !== "";
+    const systemField = req.systemPrompt === ""
+        ? {}
+        : wantsCache
+            ? {
+                system: [
+                    { type: "text", text: req.systemPrompt, cache_control: { type: "ephemeral" } },
+                ],
+            }
+            : { system: req.systemPrompt };
     return {
         model,
-        max_tokens: req.maxOutputTokens ?? DEFAULT_MAX_TOKENS,
-        ...(req.systemPrompt !== "" ? { system: req.systemPrompt } : {}),
+        max_tokens: (0, sampling_1.effectiveMaxOutputTokens)(req, DEFAULT_MAX_TOKENS),
+        ...(0, sampling_1.openAISamplingFields)(req),
+        ...systemField,
         messages: [{ role: "user", content: req.userPrompt }],
         stream: true,
-        ...(req.outputSchema !== null
-            ? {
-                tools: [
-                    {
-                        name: OUTPUT_TOOL_NAME,
-                        description: "Emit the structured output for this route.",
-                        input_schema: req.outputSchema,
-                        ...(strict ? { strict: true } : {}),
-                    },
-                ],
-                tool_choice: { type: "tool", name: OUTPUT_TOOL_NAME },
-            }
-            : {}),
+        ...toolBlock,
+    };
+}
+/**
+ * Normalize Anthropic's `message_start` usage to ModelRig's convention
+ * (research §4 fix). Anthropic's `input_tokens` is the UNCACHED portion only;
+ * ModelRig requires `tokensIn` to INCLUDE cache reads and writes as subsets, so
+ * we add them back. Pure and total — the fixture-tested seam for the subset
+ * invariant and the 5m/1h split.
+ */
+function normalizeAnthropicInputUsage(u) {
+    const uncached = u.input_tokens ?? 0;
+    const read = u.cache_read_input_tokens ?? 0;
+    const write = u.cache_creation_input_tokens ?? 0;
+    // 1h split when the extended-TTL beta reports it; clamp to the total write so
+    // a malformed payload can never make the 1h bucket exceed the write bucket.
+    const write1h = Math.min(u.cache_creation?.ephemeral_1h_input_tokens ?? 0, write);
+    return {
+        tokensIn: uncached + read + write,
+        tokensCached: read,
+        tokensCacheWrite: write,
+        tokensCacheWrite1h: write1h,
     };
 }
 function createAnthropicAdapter(config) {
     return {
         id: "anthropic",
         supports: anthropicSupports,
+        supportsTools: () => true,
         async call(req) {
             const model = req.candidate.model;
             // F1: post-stream failures still billed tokens — attach usage.
@@ -136,23 +195,47 @@ function createAnthropicAdapter(config) {
                     stream = await client.messages.create(buildAnthropicParams(model, req));
                 }
                 const events = stream;
+                // Caller-supplied tools (spec §4.4) live on the non-structured path:
+                // accumulate each `tool_use` block's input_json_delta by block index.
+                const wantTools = req.outputSchema === null && req.tools !== undefined && req.tools.length > 0;
+                const toolBlocks = new Map();
                 let text = "";
-                let promptTokens = 0;
+                // Input-side usage, normalized at message_start (subset invariant +
+                // research §4 tokensIn fix + 5m/1h split live in normalizeAnthropicInputUsage).
+                let inputUsage = {
+                    tokensIn: 0,
+                    tokensCached: 0,
+                    tokensCacheWrite: 0,
+                    tokensCacheWrite1h: 0,
+                };
                 let completionTokens = 0;
-                let cachedTokens = 0;
                 let stopReason = null;
                 for await (const event of events) {
                     if (event.type === "message_start" && event.message?.usage !== undefined) {
-                        promptTokens = event.message.usage.input_tokens ?? 0;
-                        cachedTokens = event.message.usage.cache_read_input_tokens ?? 0;
+                        inputUsage = normalizeAnthropicInputUsage(event.message.usage);
+                    }
+                    else if (event.type === "content_block_start" &&
+                        wantTools &&
+                        event.content_block?.type === "tool_use") {
+                        toolBlocks.set(event.index ?? 0, {
+                            id: event.content_block.id ?? "",
+                            name: event.content_block.name ?? "",
+                            json: "",
+                        });
                     }
                     else if (event.type === "content_block_delta") {
+                        // Tools path: input_json_delta is the tool's arguments, per block.
                         // Schema path: the forced tool's input_json_delta IS the output.
-                        // Plain path: text deltas. Never mix the two buffers.
-                        if (wantJson && event.delta?.type === "input_json_delta" && event.delta.partial_json) {
+                        // Plain path: text deltas. Never mix the buffers.
+                        if (wantTools && event.delta?.type === "input_json_delta" && event.delta.partial_json) {
+                            const blk = toolBlocks.get(event.index ?? 0);
+                            if (blk !== undefined)
+                                blk.json += event.delta.partial_json;
+                        }
+                        else if (wantJson && event.delta?.type === "input_json_delta" && event.delta.partial_json) {
                             text += event.delta.partial_json;
                         }
-                        else if (!wantJson && event.delta?.type === "text_delta" && event.delta.text) {
+                        else if (event.delta?.type === "text_delta" && event.delta.text) {
                             text += event.delta.text;
                         }
                     }
@@ -165,12 +248,23 @@ function createAnthropicAdapter(config) {
                     }
                 }
                 // Anthropic has no service tiers on this path — served is standard.
+                // The input side (tokensIn incl. cached+write, the write buckets) was
+                // normalized by normalizeAnthropicInputUsage (research §4 fix).
                 const usage = {
-                    tokensIn: promptTokens,
+                    ...inputUsage,
                     tokensOut: completionTokens,
-                    tokensCached: cachedTokens,
                     servedTier: "standard",
                 };
+                // Tool-call turn (spec §4.4): `arguments` is the raw JSON string the
+                // model streamed (OpenAI parity) — text may be empty, skip the guards.
+                if (wantTools && toolBlocks.size > 0) {
+                    const toolCalls = [...toolBlocks.values()].map((b, i) => ({
+                        id: b.id !== "" ? b.id : `call_${i}`,
+                        type: "function",
+                        function: { name: b.name, arguments: b.json === "" ? "{}" : b.json },
+                    }));
+                    return { ok: true, text, usage, toolCalls };
+                }
                 if (stopReason === "refusal") {
                     return fail("refusal", "Anthropic refused this request (stop_reason=refusal)", undefined, usage);
                 }

package/dist/adapters/build-adapters.d.ts

@@ -0,0 +1,11 @@
+/**
+ * Build the provider-adapter map from resolved config + registry facts.
+ * Extracted from index.ts (the SDK entrypoint) so that file stays under the LOC
+ * cap — the same reason wireRigArtifacts / buildExportedRouteMirrors live apart.
+ * Mechanical, one caller; behaviour is byte-identical to the prior inline block.
+ */
+import type { RigConfig } from "../config";
+import type { RegistryFacts } from "../registry/facts";
+import type { CapabilityFlag, ProviderAdapter } from "../types/provider";
+export declare function buildAdapters(config: RigConfig, registryFacts: RegistryFacts, compatStatic: ReadonlySet<CapabilityFlag>): Map<string, ProviderAdapter>;
+//# sourceMappingURL=build-adapters.d.ts.map

package/dist/adapters/build-adapters.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"build-adapters.d.ts","sourceRoot":"","sources":["../../src/adapters/build-adapters.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,WAAW,CAAC;AAC3C,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,mBAAmB,CAAC;AACvD,OAAO,KAAK,EAAE,cAAc,EAAE,eAAe,EAAE,MAAM,mBAAmB,CAAC;AAQzE,wBAAgB,aAAa,CAC3B,MAAM,EAAE,SAAS,EACjB,aAAa,EAAE,aAAa,EAC5B,YAAY,EAAE,WAAW,CAAC,cAAc,CAAC,GACxC,GAAG,CAAC,MAAM,EAAE,eAAe,CAAC,CA6B9B"}

package/dist/adapters/build-adapters.js

@@ -0,0 +1,45 @@
+"use strict";
+/**
+ * Build the provider-adapter map from resolved config + registry facts.
+ * Extracted from index.ts (the SDK entrypoint) so that file stays under the LOC
+ * cap — the same reason wireRigArtifacts / buildExportedRouteMirrors live apart.
+ * Mechanical, one caller; behaviour is byte-identical to the prior inline block.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.buildAdapters = buildAdapters;
+const anthropic_1 = require("./anthropic");
+const deepseek_1 = require("./deepseek");
+const gemini_1 = require("./gemini");
+const grok_1 = require("./grok");
+const openai_1 = require("./openai");
+const openai_compat_1 = require("./openai-compat");
+function buildAdapters(config, registryFacts, compatStatic) {
+    const adapters = new Map();
+    if (config.keys.gemini)
+        adapters.set("gemini", (0, gemini_1.createGeminiAdapter)({ apiKey: config.keys.gemini }));
+    if (config.keys.openai)
+        adapters.set("openai", (0, openai_1.createOpenAIAdapter)({ apiKey: config.keys.openai }));
+    if (config.keys.deepseek) {
+        adapters.set("deepseek", (0, deepseek_1.createDeepSeekAdapter)({ apiKey: config.keys.deepseek }));
+    }
+    if (config.keys.anthropic) {
+        adapters.set("anthropic", (0, anthropic_1.createAnthropicAdapter)({ apiKey: config.keys.anthropic }));
+    }
+    if (config.keys.grok)
+        adapters.set("grok", (0, grok_1.createGrokAdapter)({ apiKey: config.keys.grok }));
+    // A1 aggregator hosts behind the generic openai-compatible adapter.
+    for (const hostId of ["deepinfra", "fireworks"]) {
+        const apiKey = config.keys[hostId];
+        if (!apiKey)
+            continue;
+        const host = openai_compat_1.OPENAI_COMPAT_HOSTS[hostId];
+        adapters.set(hostId, (0, openai_compat_1.createOpenAICompatAdapter)({
+            providerId: hostId,
+            baseUrl: host.baseUrl,
+            apiKey,
+            ...(host.shapeModelId !== undefined ? { shapeModelId: host.shapeModelId } : {}),
+            capabilityProfile: (model) => registryFacts.flagsFor(hostId, model, compatStatic) ?? compatStatic,
+        }));
+    }
+    return adapters;
+}

package/dist/adapters/deepseek.d.ts

@@ -14,6 +14,7 @@
  */
 import type { CapabilityFlag, ProviderAdapter } from "../types/provider";
 import type { ClassifiedFailure, ShapingInput } from "./gemini";
+import type { OpenAIWireTool, OpenAIWireToolChoice } from "./tool-mapping";
 export interface DeepSeekAdapterConfig {
     readonly apiKey: string;
 }
@@ -31,9 +32,14 @@ export interface DeepSeekShapedParams {
     readonly stream_options: {
         readonly include_usage: true;
     };
+    /** CE-6 declared sampling — present only when the route/raw asks for them. */
+    readonly temperature?: number;
+    readonly top_p?: number;
     readonly response_format?: {
         readonly type: "json_object";
     };
+    readonly tools?: readonly OpenAIWireTool[];
+    readonly tool_choice?: OpenAIWireToolChoice;
 }
 export declare function buildDeepSeekParams(model: string, req: ShapingInput): DeepSeekShapedParams;
 export declare function createDeepSeekAdapter(config: DeepSeekAdapterConfig): ProviderAdapter;

package/dist/adapters/deepseek.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"deepseek.d.ts","sourceRoot":"","sources":["../../src/adapters/deepseek.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;GAaG;AAGH,OAAO,KAAK,EAIV,cAAc,EACd,eAAe,EAChB,MAAM,mBAAmB,CAAC;AAE3B,OAAO,KAAK,EAAE,iBAAiB,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAEhE,MAAM,WAAW,qBAAqB;IACpC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;CACzB;AAKD,wBAAgB,gBAAgB,CAAC,OAAO,EAAE,cAAc,EAAE,KAAK,EAAE,MAAM,GAAG,OAAO,CAMhF;AAED,wBAAgB,uBAAuB,CAAC,OAAO,EAAE,MAAM,GAAG,iBAAiB,CAiB1E;AAED,UAAU,WAAW;IACnB,QAAQ,CAAC,IAAI,EAAE,QAAQ,GAAG,MAAM,CAAC;IACjC,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,oBAAoB;IACnC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,QAAQ,EAAE,SAAS,WAAW,EAAE,CAAC;IAC1C,QAAQ,CAAC,MAAM,EAAE,IAAI,CAAC;IACtB,QAAQ,CAAC,cAAc,EAAE;QAAE,QAAQ,CAAC,aAAa,EAAE,IAAI,CAAA;KAAE,CAAC;IAC1D,QAAQ,CAAC,eAAe,CAAC,EAAE;QAAE,QAAQ,CAAC,IAAI,EAAE,aAAa,CAAA;KAAE,CAAC;CAC7D;AAOD,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,EAAE,YAAY,GAAG,oBAAoB,CAkB1F;AAED,wBAAgB,qBAAqB,CAAC,MAAM,EAAE,qBAAqB,GAAG,eAAe,CAkHpF"}
+{"version":3,"file":"deepseek.d.ts","sourceRoot":"","sources":["../../src/adapters/deepseek.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;GAaG;AAGH,OAAO,KAAK,EAIV,cAAc,EACd,eAAe,EAChB,MAAM,mBAAmB,CAAC;AAE3B,OAAO,KAAK,EAAE,iBAAiB,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAMhE,OAAO,KAAK,EAAuB,cAAc,EAAE,oBAAoB,EAAE,MAAM,gBAAgB,CAAC;AAMhG,MAAM,WAAW,qBAAqB;IACpC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;CACzB;AAKD,wBAAgB,gBAAgB,CAAC,OAAO,EAAE,cAAc,EAAE,KAAK,EAAE,MAAM,GAAG,OAAO,CAMhF;AAED,wBAAgB,uBAAuB,CAAC,OAAO,EAAE,MAAM,GAAG,iBAAiB,CAiB1E;AAED,UAAU,WAAW;IACnB,QAAQ,CAAC,IAAI,EAAE,QAAQ,GAAG,MAAM,CAAC;IACjC,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,oBAAoB;IACnC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,QAAQ,EAAE,SAAS,WAAW,EAAE,CAAC;IAC1C,QAAQ,CAAC,MAAM,EAAE,IAAI,CAAC;IACtB,QAAQ,CAAC,cAAc,EAAE;QAAE,QAAQ,CAAC,aAAa,EAAE,IAAI,CAAA;KAAE,CAAC;IAC1D,8EAA8E;IAC9E,QAAQ,CAAC,WAAW,CAAC,EAAE,MAAM,CAAC;IAC9B,QAAQ,CAAC,KAAK,CAAC,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,eAAe,CAAC,EAAE;QAAE,QAAQ,CAAC,IAAI,EAAE,aAAa,CAAA;KAAE,CAAC;IAC5D,QAAQ,CAAC,KAAK,CAAC,EAAE,SAAS,cAAc,EAAE,CAAC;IAC3C,QAAQ,CAAC,WAAW,CAAC,EAAE,oBAAoB,CAAC;CAC7C;AAOD,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,EAAE,YAAY,GAAG,oBAAoB,CA2B1F;AAED,wBAAgB,qBAAqB,CAAC,MAAM,EAAE,qBAAqB,GAAG,eAAe,CA8JpF"}

package/dist/adapters/deepseek.js

@@ -22,6 +22,9 @@ exports.classifyDeepSeekFailure = classifyDeepSeekFailure;
 exports.buildDeepSeekParams = buildDeepSeekParams;
 exports.createDeepSeekAdapter = createDeepSeekAdapter;
 const openai_1 = __importDefault(require("openai"));
+const tool_mapping_1 = require("./tool-mapping");
+const openai_stream_1 = require("./openai-stream");
+const sampling_1 = require("./sampling");
 const DEEPSEEK_BASE_URL = "https://api.deepseek.com";
 const DEFAULT_MAX_TOKENS = 64000; // R1/V3 max — full capacity for long reports (ported)
 function deepseekSupports(feature, model) {
@@ -69,17 +72,27 @@ function buildDeepSeekParams(model, req) {
     ];
     return {
         model,
-        max_tokens: req.maxOutputTokens ?? DEFAULT_MAX_TOKENS,
+        max_tokens: (0, sampling_1.effectiveMaxOutputTokens)(req, DEFAULT_MAX_TOKENS),
         messages,
         stream: true,
         stream_options: { include_usage: true },
+        ...(0, sampling_1.openAISamplingFields)(req),
         ...(wantJson ? { response_format: { type: "json_object" } } : {}),
+        ...(req.tools !== undefined && req.tools.length > 0
+            ? {
+                tools: (0, tool_mapping_1.toOpenAITools)(req.tools),
+                ...((0, tool_mapping_1.toOpenAIToolChoice)(req.toolChoice) !== undefined
+                    ? { tool_choice: (0, tool_mapping_1.toOpenAIToolChoice)(req.toolChoice) }
+                    : {}),
+            }
+            : {}),
     };
 }
 function createDeepSeekAdapter(config) {
     return {
         id: "deepseek",
         supports: deepseekSupports,
+        supportsTools: () => true,
         async call(req) {
             const model = req.candidate.model;
             // F1: post-stream failures still billed tokens — attach usage.
@@ -110,6 +123,7 @@ function createDeepSeekAdapter(config) {
                 let finishReason = null;
                 // TTFB (Phase 4 WS3): first streamed chunk = first body byte.
                 let ttfbMs;
+                const toolCallAcc = (0, tool_mapping_1.createOpenAIToolCallAccumulator)();
                 // Only delta.content is accumulated — a reasoner's reasoning_content
                 // (chain-of-thought) must never leak into the JSON buffer (ported).
                 for await (const chunk of stream) {
@@ -118,6 +132,7 @@ function createDeepSeekAdapter(config) {
                     const choice = chunk.choices[0];
                     if (choice?.delta?.content)
                         text += choice.delta.content;
+                    toolCallAcc.push(choice?.delta?.tool_calls);
                     if (choice?.finish_reason)
                         finishReason = choice.finish_reason;
                     if (chunk.usage) {
@@ -135,6 +150,11 @@ function createDeepSeekAdapter(config) {
                     servedTier: "standard",
                     ...(ttfbMs !== undefined ? { ttfbMs } : {}),
                 };
+                // Tool-call turn (spec §4.4): surface normalized calls, skip content guards.
+                const toolCalls = toolCallAcc.finalize();
+                if (toolCalls !== undefined) {
+                    return { ok: true, text, usage, toolCalls };
+                }
                 if (text.trim().length === 0) {
                     return fail("content_invalid", `empty response from DeepSeek (finish_reason=${finishReason ?? "none"})`, undefined, usage);
                 }
@@ -159,5 +179,38 @@ function createDeepSeekAdapter(config) {
                 };
             }
         },
+        async *stream(req) {
+            const model = req.candidate.model;
+            try {
+                const client = new openai_1.default({ apiKey: config.apiKey, baseURL: DEEPSEEK_BASE_URL, timeout: req.timeoutMs });
+                const params = buildDeepSeekParams(model, req);
+                const dispatchedAt = Date.now();
+                const stream = await client.chat.completions.create(params);
+                // DeepSeek: cached tokens surface as prompt_cache_hit_tokens; no tiers.
+                yield* (0, openai_stream_1.streamOpenAIDialect)(stream, req, dispatchedAt, {
+                    provider: "deepseek",
+                    cachedTokenSource: "cache_hit",
+                    mapServedTier: () => "standard",
+                });
+            }
+            catch (error) {
+                const message = error instanceof Error ? error.message : String(error);
+                const { cls, fixHint } = classifyDeepSeekFailure(message);
+                yield {
+                    kind: "final",
+                    result: {
+                        ok: false,
+                        failure: {
+                            class: cls,
+                            message,
+                            provider: "deepseek",
+                            model,
+                            cause: error,
+                            ...(fixHint !== undefined ? { fixHint } : {}),
+                        },
+                    },
+                };
+            }
+        },
     };
 }

package/dist/adapters/gemini.d.ts

@@ -33,6 +33,16 @@ export declare function classifyGeminiFailure(message: string): ClassifiedFailur
 /** The candidate-independent part of an AdapterRequest (also what the pure
  * shaping helpers take, so unit tests never need a branded CandidateRef). */
 export type ShapingInput = Omit<AdapterRequest, "candidate">;
+/**
+ * CE-2 F3 fail-closed guard: `reasoning` declared on a Gemini model that is NOT
+ * thinking-capable would 400 at the provider (thinkingConfig on a 2.x model is
+ * rejected). Pre-empt with a classified TERMINAL failure so the declaration is
+ * never silently dropped and never turns into an unguarded provider 400 — the
+ * same fail-closed discipline as the raw-lane provider guard (spec §3 F1/F3).
+ * Returns the failure class + fixHint, or null when the request is fine (no
+ * reasoning declared, or a thinking-capable model). Pure — unit-tested directly.
+ */
+export declare function geminiReasoningUnsupported(model: string, req: ShapingInput): ClassifiedFailure | null;
 export interface GeminiShapedRequest {
     readonly model: string;
     readonly contents: string;

package/dist/adapters/gemini.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"gemini.d.ts","sourceRoot":"","sources":["../../src/adapters/gemini.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;GAcG;AAGH,OAAO,KAAK,EACV,cAAc,EAGd,cAAc,EACd,eAAe,EAEhB,MAAM,mBAAmB,CAAC;AAC3B,OAAO,KAAK,EAAE,YAAY,EAAc,MAAM,mBAAmB,CAAC;AAElE,MAAM,WAAW,mBAAmB;IAClC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;CACzB;AAyBD,wBAAgB,cAAc,CAAC,OAAO,EAAE,cAAc,EAAE,KAAK,EAAE,MAAM,GAAG,OAAO,CAG9E;AAED,MAAM,WAAW,iBAAiB;IAChC,QAAQ,CAAC,GAAG,EAAE,YAAY,CAAC;IAC3B,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC;CAC3B;AAED;;;;;GAKG;AACH,wBAAgB,qBAAqB,CAAC,OAAO,EAAE,MAAM,GAAG,iBAAiB,CAyCxE;AAED;6EAC6E;AAC7E,MAAM,MAAM,YAAY,GAAG,IAAI,CAAC,cAAc,EAAE,WAAW,CAAC,CAAC;AAE7D,MAAM,WAAW,mBAAmB;IAClC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAC1C;AAED,iEAAiE;AACjE,wBAAgB,kBAAkB,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,EAAE,YAAY,GAAG,mBAAmB,CAkCxF;AA0BD,wBAAgB,mBAAmB,CAAC,MAAM,EAAE,mBAAmB,GAAG,eAAe,CAyHhF"}
+{"version":3,"file":"gemini.d.ts","sourceRoot":"","sources":["../../src/adapters/gemini.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;GAcG;AAGH,OAAO,KAAK,EACV,cAAc,EAGd,cAAc,EACd,eAAe,EAEhB,MAAM,mBAAmB,CAAC;AAC3B,OAAO,KAAK,EAAE,YAAY,EAAc,MAAM,mBAAmB,CAAC;AAKlE,MAAM,WAAW,mBAAmB;IAClC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;CACzB;AAkCD,wBAAgB,cAAc,CAAC,OAAO,EAAE,cAAc,EAAE,KAAK,EAAE,MAAM,GAAG,OAAO,CAG9E;AAED,MAAM,WAAW,iBAAiB;IAChC,QAAQ,CAAC,GAAG,EAAE,YAAY,CAAC;IAC3B,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC;CAC3B;AAED;;;;;GAKG;AACH,wBAAgB,qBAAqB,CAAC,OAAO,EAAE,MAAM,GAAG,iBAAiB,CAyCxE;AAED;6EAC6E;AAC7E,MAAM,MAAM,YAAY,GAAG,IAAI,CAAC,cAAc,EAAE,WAAW,CAAC,CAAC;AAE7D;;;;;;;;GAQG;AACH,wBAAgB,0BAA0B,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,EAAE,YAAY,GAAG,iBAAiB,GAAG,IAAI,CAQrG;AAED,MAAM,WAAW,mBAAmB;IAClC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAC1C;AAED,iEAAiE;AACjE,wBAAgB,kBAAkB,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,EAAE,YAAY,GAAG,mBAAmB,CAgDxF;AA0BD,wBAAgB,mBAAmB,CAAC,MAAM,EAAE,mBAAmB,GAAG,eAAe,CA+JhF"}

package/dist/adapters/gemini.js

@@ -17,9 +17,12 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.geminiSupports = geminiSupports;
 exports.classifyGeminiFailure = classifyGeminiFailure;
+exports.geminiReasoningUnsupported = geminiReasoningUnsupported;
 exports.buildGeminiRequest = buildGeminiRequest;
 exports.createGeminiAdapter = createGeminiAdapter;
 const genai_1 = require("@google/genai");
+const tool_mapping_1 = require("./tool-mapping");
+const sampling_1 = require("./sampling");
 /** Models that accept thinkingConfig — passing it to 2.x models 400s.
  * Ported from the production THINKING_CAPABLE_MODELS set. */
 const THINKING_CAPABLE_MODELS = new Set([
@@ -27,7 +30,16 @@ const THINKING_CAPABLE_MODELS = new Set([
     "gemini-3.0-pro-preview",
     "gemini-3.1-flash-preview",
 ]);
-/** Gemini 3 requires temperature=1.0 — lower values cause looping (ported). */
+/** Gemini-3 default. Google STRONGLY recommends temperature=1.0 for all
+ * Gemini 3 models and warns that values below 1.0 can cause looping / degraded
+ * output — concentrated on complex reasoning/math tasks, not simple
+ * classification (verified 2026-08-20 against ai.google.dev/gemini-api/docs/gemini-3;
+ * the source pipeline ran gemini-3.1-pro at 0.1 for CLASSIFICATION without
+ * looping, which the reasoning-task carve-out explains). It is a recommendation,
+ * not a hard API cap: the accepted range stays 0–2. This is the DEFAULT the
+ * adapter sends when a route declares no `sampling.temperature`; a declared
+ * value overrides it and is sent as-is (CE-6), with a load-time warning emitted
+ * for a sub-1.0 value on a Gemini-3 candidate (routes/loader.ts). */
 const GEMINI_TEMPERATURE = 1.0;
 const DEFAULT_MAX_OUTPUT_TOKENS = 64000; // thinking tokens count against this
 const TOP_P = 0.95;
@@ -88,15 +100,41 @@ function classifyGeminiFailure(message) {
     }
     return { cls: "network" };
 }
+/**
+ * CE-2 F3 fail-closed guard: `reasoning` declared on a Gemini model that is NOT
+ * thinking-capable would 400 at the provider (thinkingConfig on a 2.x model is
+ * rejected). Pre-empt with a classified TERMINAL failure so the declaration is
+ * never silently dropped and never turns into an unguarded provider 400 — the
+ * same fail-closed discipline as the raw-lane provider guard (spec §3 F1/F3).
+ * Returns the failure class + fixHint, or null when the request is fine (no
+ * reasoning declared, or a thinking-capable model). Pure — unit-tested directly.
+ */
+function geminiReasoningUnsupported(model, req) {
+    if (req.reasoning !== undefined && !THINKING_CAPABLE_MODELS.has(model)) {
+        return {
+            cls: "invariant_violation",
+            fixHint: `reasoning control requires a thinking-capable Gemini model; "${model}" is not one — remove reasoning or choose a thinking-capable model`,
+        };
+    }
+    return null;
+}
 /** Pure request shaping — everything except the abort signal. */
 function buildGeminiRequest(model, req) {
     const usingCache = req.cache?.key !== undefined && req.cache.key !== "";
+    // CE-6: a declared sampling field OVERRIDES the ported constant; absent, the
+    // constant stands (byte-identical to pre-CE-6). topK stays internal (gemini-only,
+    // not on the customer surface — spec §1 non-goals).
+    // CE-2 F3: a declared reasoning level replaces the hardcoded `"high"` for the
+    // thinking-capable models; absent, `"high"` stands (byte-identical to
+    // pre-CE-2). A declared level on a NON-capable model is rejected fail-closed in
+    // call() (geminiReasoningUnsupported) — it never reaches this shaping.
+    const thinkingLevel = req.reasoning?.level ?? "high";
     const config = {
-        temperature: GEMINI_TEMPERATURE,
-        maxOutputTokens: req.maxOutputTokens ?? DEFAULT_MAX_OUTPUT_TOKENS,
-        topP: TOP_P,
+        temperature: req.sampling?.temperature ?? GEMINI_TEMPERATURE,
+        maxOutputTokens: (0, sampling_1.effectiveMaxOutputTokens)(req, DEFAULT_MAX_OUTPUT_TOKENS),
+        topP: req.sampling?.topP ?? TOP_P,
         topK: TOP_K,
-        ...(THINKING_CAPABLE_MODELS.has(model) ? { thinkingConfig: { thinkingLevel: "high" } } : {}),
+        ...(THINKING_CAPABLE_MODELS.has(model) ? { thinkingConfig: { thinkingLevel } } : {}),
     };
     if (req.outputSchema !== null) {
         config.responseMimeType = "application/json";
@@ -115,6 +153,14 @@ function buildGeminiRequest(model, req) {
         if (req.grounding === true) {
             config.tools = [{ googleSearch: {} }];
         }
+        else if (req.tools !== undefined && req.tools.length > 0) {
+            // Caller-supplied tools (spec §4.4) → Gemini functionDeclarations.
+            // Grounding wins the single `tools` slot when both are asked for.
+            config.tools = (0, tool_mapping_1.toGeminiTools)(req.tools);
+            const toolConfig = (0, tool_mapping_1.toGeminiToolConfig)(req.toolChoice);
+            if (toolConfig !== undefined)
+                config.toolConfig = toolConfig;
+        }
     }
     if (req.requestedTier === "flex") {
         config.serviceTier = "flex";
@@ -139,6 +185,7 @@ function createGeminiAdapter(config) {
     return {
         id: "gemini",
         supports: geminiSupports,
+        supportsTools: () => true,
         async call(req) {
             const model = req.candidate.model;
             const abortController = new AbortController();
@@ -156,6 +203,13 @@ function createGeminiAdapter(config) {
                 },
                 ...(usage !== undefined ? { usage } : {}),
             });
+            // CE-2 F3: reject reasoning on a non-thinking model BEFORE any dispatch —
+            // classified, terminal, never a silent drop (spec §3 F3).
+            const reasoningUnsupported = geminiReasoningUnsupported(model, req);
+            if (reasoningUnsupported !== null) {
+                clearTimeout(timeoutId);
+                return fail(reasoningUnsupported.cls, `reasoning control declared on the non-thinking Gemini model "${model}"`, reasoningUnsupported.fixHint);
+            }
             try {
                 const shaped = buildGeminiRequest(model, req);
                 const dispatchedAt = Date.now();
@@ -176,9 +230,26 @@ function createGeminiAdapter(config) {
                 const tokensOut = (usageMetadata.candidatesTokenCount ?? 0) + (usageMetadata.thoughtsTokenCount ?? 0);
                 const tokensCached = usageMetadata.cachedContentTokenCount ?? 0;
                 const usage = { tokensIn, tokensOut, tokensCached, servedTier, ttfbMs };
-                const candidates = response
-                    .candidates;
+                const candidates = response.candidates;
                 const finishReason = candidates?.[0]?.finishReason;
+                // Tool-call turn (spec §4.4): collect functionCall parts. When present,
+                // surface normalized calls and skip the text/empty/truncation guards.
+                // Read any accompanying text from the parts (not the `response.text`
+                // getter, which can throw/warn on a functionCall-only response).
+                if (req.tools !== undefined && req.tools.length > 0) {
+                    const parts = candidates?.[0]?.content?.parts ?? [];
+                    const fnCalls = parts
+                        .map((p) => p.functionCall)
+                        .filter((fc) => fc !== undefined && fc !== null);
+                    const toolCalls = (0, tool_mapping_1.geminiFunctionCallsToToolCalls)(fnCalls);
+                    if (toolCalls !== undefined) {
+                        const toolText = parts
+                            .map((p) => p.text)
+                            .filter((t) => typeof t === "string")
+                            .join("");
+                        return { ok: true, text: toolText, usage, toolCalls };
+                    }
+                }
                 if (finishReason === "SAFETY" || finishReason === "PROHIBITED_CONTENT") {
                     return fail("refusal", `Gemini blocked output via safety filter (finishReason=${finishReason})`, undefined, usage);
                 }

package/dist/adapters/grok.d.ts

@@ -12,6 +12,7 @@
  */
 import type { CapabilityFlag, ProviderAdapter } from "../types/provider";
 import type { ClassifiedFailure, ShapingInput } from "./gemini";
+import type { OpenAIWireTool, OpenAIWireToolChoice } from "./tool-mapping";
 export interface GrokAdapterConfig {
     readonly apiKey: string;
 }
@@ -29,6 +30,12 @@ export interface GrokShapedParams {
     readonly stream_options: {
         readonly include_usage: true;
     };
+    /** CE-6 declared sampling — present only when the route/raw asks for them. */
+    readonly temperature?: number;
+    readonly top_p?: number;
+    /** Prompt-cache routing hint (WS-C) — the stable-prefix fingerprint the run
+     * loop stamps; xAI keeps a warm cache keyed to it (writes free today). */
+    readonly prompt_cache_key?: string;
     readonly response_format?: {
         readonly type: "json_schema";
         readonly json_schema: {
@@ -37,6 +44,8 @@ export interface GrokShapedParams {
             readonly strict: true;
         };
     };
+    readonly tools?: readonly OpenAIWireTool[];
+    readonly tool_choice?: OpenAIWireToolChoice;
 }
 export declare function buildGrokParams(model: string, req: ShapingInput): GrokShapedParams;
 export declare function createGrokAdapter(config: GrokAdapterConfig): ProviderAdapter;

package/dist/adapters/grok.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"grok.d.ts","sourceRoot":"","sources":["../../src/adapters/grok.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAGH,OAAO,KAAK,EAIV,cAAc,EACd,eAAe,EAChB,MAAM,mBAAmB,CAAC;AAE3B,OAAO,KAAK,EAAE,iBAAiB,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAEhE,MAAM,WAAW,iBAAiB;IAChC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;CACzB;AAKD,wBAAgB,YAAY,CAAC,OAAO,EAAE,cAAc,EAAE,MAAM,EAAE,MAAM,GAAG,OAAO,CAa7E;AAED,wBAAgB,mBAAmB,CAAC,OAAO,EAAE,MAAM,GAAG,iBAAiB,CAsBtE;AAED,UAAU,WAAW;IACnB,QAAQ,CAAC,IAAI,EAAE,QAAQ,GAAG,MAAM,CAAC;IACjC,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,gBAAgB;IAC/B,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,QAAQ,EAAE,SAAS,WAAW,EAAE,CAAC;IAC1C,QAAQ,CAAC,MAAM,EAAE,IAAI,CAAC;IACtB,QAAQ,CAAC,cAAc,EAAE;QAAE,QAAQ,CAAC,aAAa,EAAE,IAAI,CAAA;KAAE,CAAC;IAC1D,QAAQ,CAAC,eAAe,CAAC,EAAE;QACzB,QAAQ,CAAC,IAAI,EAAE,aAAa,CAAC;QAC7B,QAAQ,CAAC,WAAW,EAAE;YAAE,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;YAAC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;YAAC,QAAQ,CAAC,MAAM,EAAE,IAAI,CAAA;SAAE,CAAC;KACjG,CAAC;CACH;AAED,wBAAgB,eAAe,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,EAAE,YAAY,GAAG,gBAAgB,CAwBlF;AAED,wBAAgB,iBAAiB,CAAC,MAAM,EAAE,iBAAiB,GAAG,eAAe,CA4G5E"}
+{"version":3,"file":"grok.d.ts","sourceRoot":"","sources":["../../src/adapters/grok.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAGH,OAAO,KAAK,EAIV,cAAc,EACd,eAAe,EAChB,MAAM,mBAAmB,CAAC;AAE3B,OAAO,KAAK,EAAE,iBAAiB,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAMhE,OAAO,KAAK,EAAuB,cAAc,EAAE,oBAAoB,EAAE,MAAM,gBAAgB,CAAC;AAMhG,MAAM,WAAW,iBAAiB;IAChC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;CACzB;AAKD,wBAAgB,YAAY,CAAC,OAAO,EAAE,cAAc,EAAE,MAAM,EAAE,MAAM,GAAG,OAAO,CAc7E;AAED,wBAAgB,mBAAmB,CAAC,OAAO,EAAE,MAAM,GAAG,iBAAiB,CAsBtE;AAED,UAAU,WAAW;IACnB,QAAQ,CAAC,IAAI,EAAE,QAAQ,GAAG,MAAM,CAAC;IACjC,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,gBAAgB;IAC/B,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,QAAQ,EAAE,SAAS,WAAW,EAAE,CAAC;IAC1C,QAAQ,CAAC,MAAM,EAAE,IAAI,CAAC;IACtB,QAAQ,CAAC,cAAc,EAAE;QAAE,QAAQ,CAAC,aAAa,EAAE,IAAI,CAAA;KAAE,CAAC;IAC1D,8EAA8E;IAC9E,QAAQ,CAAC,WAAW,CAAC,EAAE,MAAM,CAAC;IAC9B,QAAQ,CAAC,KAAK,CAAC,EAAE,MAAM,CAAC;IACxB;8EAC0E;IAC1E,QAAQ,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC;IACnC,QAAQ,CAAC,eAAe,CAAC,EAAE;QACzB,QAAQ,CAAC,IAAI,EAAE,aAAa,CAAC;QAC7B,QAAQ,CAAC,WAAW,EAAE;YAAE,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;YAAC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;YAAC,QAAQ,CAAC,MAAM,EAAE,IAAI,CAAA;SAAE,CAAC;KACjG,CAAC;IACF,QAAQ,CAAC,KAAK,CAAC,EAAE,SAAS,cAAc,EAAE,CAAC;IAC3C,QAAQ,CAAC,WAAW,CAAC,EAAE,oBAAoB,CAAC;CAC7C;AAED,wBAAgB,eAAe,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,EAAE,YAAY,GAAG,gBAAgB,CAsClF;AAED,wBAAgB,iBAAiB,CAAC,MAAM,EAAE,iBAAiB,GAAG,eAAe,CAgK5E"}