pi-fireworks-provider 1.0.2 → 1.1.0

package/README.md

@@ -4,7 +4,7 @@
 
 **31+ models through [Fireworks AI](https://fireworks.ai/)**
 
-_Kimi, MiniMax, GLM, DeepSeek, GPT-OSS — unified OpenAI-compatible API for [pi](https://github.com/earendil-works/pi-coding-agent)._
+_Kimi, MiniMax, GLM, DeepSeek, GPT-OSS — via Fireworks AI's Anthropic Messages and OpenAI-compatible endpoints for [pi](https://github.com/earendil-works/pi-coding-agent)._
 
 [![pi extension](https://img.shields.io/badge/pi-extension-blueviolet)](https://github.com/earendil-works/pi-coding-agent)
 [![license](https://img.shields.io/badge/license-MIT-blue)](./LICENSE)
@@ -16,7 +16,8 @@ _Kimi, MiniMax, GLM, DeepSeek, GPT-OSS — unified OpenAI-compatible API for [pi
 ## Features
 
 - **35+ AI Models** including Kimi K2.5, MiniMax M2.5, GLM 4.5/4.7/5, DeepSeek V3.1/V3.2, DeepSeek V4 Flash, and GPT-OSS
-- **Unified API** via Fireworks AI's OpenAI-compatible completions endpoint
+- **Dual API support** via Fireworks AI's Anthropic Messages and OpenAI-compatible completions endpoints (per-model routing, matching pi core's Fireworks provider)
+- **Service tiers** — toggle Fireworks `priority` vs `standard` per request on supported models (with priority pricing reflected in cost tracking), via a keybinding, `/fireworks-tier`, and a footer status area
 - **Cost Tracking** with per-model pricing for budget management
 - **Reasoning Models** support for advanced reasoning capabilities
 - **Vision Support** for image-capable models
@@ -105,6 +106,50 @@ pi
 | Qwen3 VL 30B A3B Thinking | Text + Image | 262K | 0 | Free | Free |
 *Costs are per million tokens. Prices subject to change - check [fireworks.ai](https://fireworks.ai) for current pricing.*
 
+## Service Tiers
+
+Fireworks exposes a `service_tier` request field (`standard` | `priority`) on its chat-completions endpoint. The **priority** tier trades higher per-token pricing for higher throughput / lower latency. This is orthogonal to the `-fast`/`-turbo` router model IDs (which are separate models) — service tiers apply to the base models below.
+
+| Model | Priority Uncached Input | Priority Cached Input | Priority Output |
+| --- | --- | --- | --- |
+| GLM 5.2 | $1.75/M | $0.175/M | $5.5/M |
+| Kimi K2.7 Code | $1.43/M | $0.29/M | $6/M |
+| Minimax M3 | $0.45/M | $0.09/M | $1.8/M |
+| DeepSeek V4 Pro | $2.61/M | $0.218/M | $5.22/M |
+| Kimi K2.6 | $1.5/M | $0.22/M | $6/M |
+| MiniMax M2.7 | $0.45/M | $0.09/M | $1.8/M |
+| GLM 5.1 | $2.1/M | $0.39/M | $6.6/M |
+| GPT OSS 120B | $0.18/M | $0.018/M | $0.72/M |
+| DeepSeek V4 Flash | $0.21/M | $0.045/M | $0.42/M |
+
+*Priority pricing is roughly 1.2–1.5× the standard rate. `cacheWrite` is not tiered.*
+
+**Switching tiers:**
+
+- **Keybinding:** `ctrl+shift+l` (default) toggles `standard` ↔ `priority` for the active supported model. No-op with an info notice for unsupported models.
+- **Command:** `/fireworks-tier standard|priority|toggle`.
+- **Status area:** a dim `tier: standard` / `tier: ⚡priority` line is shown in the footer for supported models while a Fireworks model is active.
+
+The selection is persisted per session (survives `/reload` and resume). When `priority` is active, `service_tier: "priority"` is injected into every request and finalized cost is recomputed against the priority rates above.
+
+**Configuration** — `~/.pi/agent/extensions/fireworks.json` (created with defaults on first load):
+
+```json
+{
+  "serviceTier": {
+    "default": "standard",
+    "keybinding": "ctrl+shift+l",
+    "display": "statusbar"
+  }
+}
+```
+
+- `default` — tier used until you toggle (`standard` | `priority`).
+- `keybinding` — any [pi key format](https://github.com/earendil-works/pi-coding-agent/blob/main/docs/keybindings.md) (e.g. `ctrl+shift+l`, `ctrl+shift+k`). Requires `/reload` after changing. On macOS browser terminals (localterm), avoid `alt`/`ctrl+alt` (Option produces special chars) and `ctrl+shift+t/w/n/c/v` (browser/localterm tab + copy/paste shortcuts).
+- `display` — `statusbar` (footer status area) or `off` (hide the tier indicator).
+
+> **Note:** The OpenAI completions endpoint accepts `service_tier` directly (per Fireworks' API). The Anthropic Messages endpoint passes the top-level field through as an extra. If a supported Anthropic-routed model rejects it, file an issue so we can gate injection by API.
+
 ## Usage
 
 After loading the extension, use the `/model` command in pi to select your preferred model:

package/custom-models.json

@@ -147,5 +147,72 @@
     ],
     "contextWindow": 262000,
     "maxTokens": 0
+  },
+  {
+    "id": "accounts/fireworks/models/qwen3p7-plus",
+    "name": "Qwen 3.7 Plus",
+    "reasoning": true,
+    "cost": {
+      "input": 0.4,
+      "output": 1.6,
+      "cacheRead": 0.08,
+      "cacheWrite": 0
+    },
+    "input": [
+      "text",
+      "image"
+    ],
+    "contextWindow": 262144,
+    "maxTokens": 65536
+  },
+  {
+    "id": "accounts/fireworks/routers/glm-5p2-fast",
+    "name": "GLM 5.2 Fast",
+    "reasoning": true,
+    "cost": {
+      "input": 2.1,
+      "output": 6.6,
+      "cacheRead": 0.21,
+      "cacheWrite": 0
+    },
+    "input": [
+      "text"
+    ],
+    "contextWindow": 1048575,
+    "maxTokens": 131072
+  },
+  {
+    "id": "accounts/fireworks/routers/kimi-k2p6-fast",
+    "name": "Kimi K2.6 Fast",
+    "reasoning": true,
+    "cost": {
+      "input": 2,
+      "output": 8,
+      "cacheRead": 0.3,
+      "cacheWrite": 0
+    },
+    "input": [
+      "text",
+      "image"
+    ],
+    "contextWindow": 262000,
+    "maxTokens": 262000
+  },
+  {
+    "id": "accounts/fireworks/routers/kimi-k2p7-code-fast",
+    "name": "Kimi K2.7 Code Fast",
+    "reasoning": true,
+    "cost": {
+      "input": 1.9,
+      "output": 8,
+      "cacheRead": 0.38,
+      "cacheWrite": 0
+    },
+    "input": [
+      "text",
+      "image"
+    ],
+    "contextWindow": 262000,
+    "maxTokens": 262000
   }
 ]

package/index.ts

@@ -34,10 +34,15 @@ import path from "path";
 
 // ─── Types ────────────────────────────────────────────────────────────────────
 
+type FireworksApi = "anthropic-messages" | "openai-completions";
+
 interface JsonModel {
   id: string;
   name: string;
+  api?: FireworksApi;
+  baseUrl?: string;
   reasoning: boolean;
+  thinkingLevelMap?: Record<string, string | null>;
   input: string[];
   cost: {
     input: number;
@@ -47,18 +52,17 @@ interface JsonModel {
   };
   contextWindow: number;
   maxTokens: number;
-  compat?: {
-    supportsDeveloperRole?: boolean;
-    supportsStore?: boolean;
-    maxTokensField?: "max_completion_tokens" | "max_tokens";
-    thinkingFormat?: "openai" | "zai" | "qwen" | "qwen-chat-template";
-    supportsReasoningEffort?: boolean;
-  };
+  // Loose on purpose: shape depends on `api` (OpenAICompletionsCompat vs
+  // AnthropicMessagesCompat). pi-ai validates at runtime.
+  compat?: Record<string, unknown>;
 }
 
 interface PatchEntry {
   name?: string;
+  api?: FireworksApi;
+  baseUrl?: string;
   reasoning?: boolean;
+  thinkingLevelMap?: Record<string, string | null>;
   input?: string[];
   cost?: {
     input?: number;
@@ -79,7 +83,10 @@ function applyPatch(model: JsonModel, patch: PatchEntry): JsonModel {
   const result = { ...model };
 
   if (patch.name !== undefined) result.name = patch.name;
+  if (patch.api !== undefined) result.api = patch.api;
+  if (patch.baseUrl !== undefined) result.baseUrl = patch.baseUrl;
   if (patch.reasoning !== undefined) result.reasoning = patch.reasoning;
+  if (patch.thinkingLevelMap !== undefined) result.thinkingLevelMap = patch.thinkingLevelMap;
   if (patch.input !== undefined) result.input = patch.input;
   if (patch.contextWindow !== undefined) result.contextWindow = patch.contextWindow;
   if (patch.maxTokens !== undefined) result.maxTokens = patch.maxTokens;
@@ -335,9 +342,176 @@ function stripAnchorBleedInPlace(obj: Record<string, unknown>): void {
   }
 }
 
+// ─── Service Tier (standard / priority) ──────────────────────────────────────
+
+// Fireworks exposes a `service_tier` request field ("standard" | "priority") on
+// its chat-completions endpoint. The priority tier trades higher per-token
+// pricing for higher throughput / lower latency on supported models. This is
+// orthogonal to the "fast" router model IDs (e.g. routers/...-fast), which are
+// separate models; priority applies to the base models below.
+//
+// Per-request priority pricing (USD per million tokens) from Fireworks' tier
+// reference. cacheWrite is not tiered (stays 0).
+const PRIORITY_PRICING: Record<string, { input: number; output: number; cacheRead: number; cacheWrite: number }> = {
+  "accounts/fireworks/models/glm-5p2":          { input: 1.75, output: 5.5,  cacheRead: 0.175, cacheWrite: 0 },
+  "accounts/fireworks/models/kimi-k2p7-code":   { input: 1.43, output: 6,   cacheRead: 0.29,  cacheWrite: 0 },
+  "accounts/fireworks/models/minimax-m3":       { input: 0.45, output: 1.8,  cacheRead: 0.09,  cacheWrite: 0 },
+  "accounts/fireworks/models/deepseek-v4-pro":  { input: 2.61, output: 5.22, cacheRead: 0.218, cacheWrite: 0 },
+  "accounts/fireworks/models/kimi-k2p6":        { input: 1.5,  output: 6,   cacheRead: 0.22,  cacheWrite: 0 },
+  "accounts/fireworks/models/minimax-m2p7":    { input: 0.45, output: 1.8,  cacheRead: 0.09,  cacheWrite: 0 },
+  "accounts/fireworks/models/glm-5p1":         { input: 2.1,  output: 6.6,  cacheRead: 0.39,  cacheWrite: 0 },
+  "accounts/fireworks/models/gpt-oss-120b":     { input: 0.18, output: 0.72, cacheRead: 0.018, cacheWrite: 0 },
+  "accounts/fireworks/models/deepseek-v4-flash":{ input: 0.21, output: 0.42, cacheRead: 0.045, cacheWrite: 0 },
+};
+
+type ServiceTier = "standard" | "priority";
+
+interface ServiceTierConfig {
+  default: ServiceTier;
+  keybinding: string;
+  display: "statusbar" | "off";
+}
+
+interface FireworksConfig {
+  serviceTier: ServiceTierConfig;
+}
+
+const FIREWORKS_CONFIG_PATH = path.join(getAgentDir(), "extensions", "fireworks.json");
+const TIER_ENTRY_TYPE = "fireworks-service-tier";
+const TIER_STATUS_KEY = "fireworks-tier";
+const DEFAULT_SERVICE_TIER_CONFIG: ServiceTierConfig = {
+  default: "standard",
+  keybinding: "ctrl+shift+l",
+  display: "statusbar",
+};
+const DEFAULT_FIREWORKS_CONFIG: FireworksConfig = { serviceTier: DEFAULT_SERVICE_TIER_CONFIG };
+
+function isValidTier(v: unknown): v is ServiceTier {
+  return v === "standard" || v === "priority";
+}
+
+function loadFireworksConfig(): FireworksConfig {
+  try {
+    const raw = JSON.parse(fs.readFileSync(FIREWORKS_CONFIG_PATH, "utf8"));
+    const st = raw?.serviceTier ?? {};
+    return {
+      serviceTier: {
+        default: isValidTier(st.default) ? st.default : DEFAULT_SERVICE_TIER_CONFIG.default,
+        keybinding: typeof st.keybinding === "string" && st.keybinding.length > 0 ? st.keybinding : DEFAULT_SERVICE_TIER_CONFIG.keybinding,
+        display: st.display === "off" ? "off" : "statusbar",
+      },
+    };
+  } catch {
+    // Config missing or invalid — write defaults so the user can discover it.
+    try {
+      fs.mkdirSync(path.dirname(FIREWORKS_CONFIG_PATH), { recursive: true });
+      fs.writeFileSync(FIREWORKS_CONFIG_PATH, JSON.stringify(DEFAULT_FIREWORKS_CONFIG, null, 2) + "\n");
+    } catch {
+      // Write failure is non-fatal — defaults still work in memory.
+    }
+    return { serviceTier: { ...DEFAULT_SERVICE_TIER_CONFIG } };
+  }
+}
+
+let fireworksConfig = loadFireworksConfig();
+
+// Held so module-scope helpers (setTier) can call pi.appendEntry.
+let piRef: ExtensionAPI | null = null;
+
+function isPriorityApplicable(id: string | undefined): boolean {
+  return !!id && Object.prototype.hasOwnProperty.call(PRIORITY_PRICING, id);
+}
+
+// Session state: the active service tier. Replayed from session entries on
+// session_start so it survives reload / resume.
+let currentTier: ServiceTier = fireworksConfig.serviceTier.default;
+
+function replayTierState(ctx: any, defaultTier: ServiceTier): void {
+  let tier: ServiceTier = defaultTier;
+  for (const entry of ctx.sessionManager.getBranch()) {
+    if (entry?.type === "custom" && entry.customType === TIER_ENTRY_TYPE && entry.data) {
+      const t = entry.data.tier;
+      if (isValidTier(t)) tier = t;
+    }
+  }
+  currentTier = tier;
+}
+
+function updateTierStatus(ctx: any): void {
+  if (fireworksConfig.serviceTier.display === "off") {
+    ctx.ui.setStatus(TIER_STATUS_KEY, undefined);
+    return;
+  }
+  let model: any;
+  try {
+    model = ctx.model;
+  } catch {
+    model = undefined;
+  }
+  if (!model || model.provider !== "fireworks" || !isPriorityApplicable(model.id)) {
+    ctx.ui.setStatus(TIER_STATUS_KEY, undefined);
+    return;
+  }
+  const label = currentTier === "priority" ? "tier: ⚡priority" : "tier: standard";
+  try {
+    ctx.ui.setStatus(TIER_STATUS_KEY, ctx.ui.theme.fg("dim", label));
+  } catch {
+    // setStatus / theme are no-ops without a UI runner.
+  }
+}
+
+function setTier(ctx: any, tier: ServiceTier): void {
+  currentTier = tier;
+  try {
+    piRef?.appendEntry(TIER_ENTRY_TYPE, { tier });
+  } catch {
+    // appendEntry outside a session context is non-fatal.
+  }
+  updateTierStatus(ctx);
+}
+
+function toggleTier(ctx: any): void {
+  let model: any;
+  try {
+    model = ctx.model;
+  } catch {
+    model = undefined;
+  }
+  if (!model || model.provider !== "fireworks") {
+    try { ctx.ui.notify("Fireworks service tier only applies to Fireworks models.", "info"); } catch {}
+    return;
+  }
+  if (!isPriorityApplicable(model.id)) {
+    try { ctx.ui.notify(`Service tier is not available for ${model.name || model.id}.`, "info"); } catch {}
+    return;
+  }
+  const next: ServiceTier = currentTier === "priority" ? "standard" : "priority";
+  setTier(ctx, next);
+  try { ctx.ui.notify(`Fireworks service tier: ${next}`, "info"); } catch {}
+}
+
+// Recompute a finalized assistant message's cost against priority pricing so
+// cost tracking reflects the premium tier instead of the model's base cost.
+function recomputePriorityCost(message: any): any {
+  const pricing = message?.model ? PRIORITY_PRICING[message.model] : undefined;
+  if (!pricing) return undefined;
+  const usage = message?.usage;
+  if (!usage) return undefined;
+  const cost = {
+    input: (pricing.input / 1_000_000) * (usage.input || 0),
+    output: (pricing.output / 1_000_000) * (usage.output || 0),
+    cacheRead: (pricing.cacheRead / 1_000_000) * (usage.cacheRead || 0),
+    cacheWrite: 0,
+    total: 0,
+  };
+  cost.total = cost.input + cost.output + cost.cacheRead + cost.cacheWrite;
+  return { ...message, usage: { ...usage, cost } };
+}
+
 // ─── Extension Entry Point ────────────────────────────────────────────────────
 
 export default function (pi: ExtensionAPI) {
+  piRef = pi;
   const embeddedModels = modelsData as JsonModel[];
   const customModels = customModelsData as JsonModel[];
   const patches = patchData as PatchData;
@@ -356,6 +530,9 @@ export default function (pi: ExtensionAPI) {
     revalidateAbort?.abort();
     revalidateAbort = new AbortController();
     const signal = revalidateAbort.signal;
+    fireworksConfig = loadFireworksConfig();
+    replayTierState(ctx, fireworksConfig.serviceTier.default);
+    updateTierStatus(ctx);
     resolveApiKey(ctx.modelRegistry).then(() => {
       revalidateModels(cachedApiKey, embeddedModels, signal).then((freshBase) => {
         if (freshBase && !signal.aborted) {
@@ -370,8 +547,9 @@ export default function (pi: ExtensionAPI) {
     });
   });
 
-  pi.on("session_shutdown", () => {
+  pi.on("session_shutdown", (_event, ctx) => {
     revalidateAbort?.abort();
+    try { ctx.ui.setStatus(TIER_STATUS_KEY, undefined); } catch {}
   });
 
   // Sanitize JSON Schema patterns for Kimi models before sending to Fireworks.
@@ -380,34 +558,52 @@ export default function (pi: ExtensionAPI) {
   // present. We strip anchors from simple patterns and drop patterns that
   // combine alternation with anchors entirely.
   pi.on("before_provider_request", (event, ctx) => {
-    if (!isFireworksKimiModel(ctx.model)) return;
+    const model = ctx.model;
+    if (!model || model.provider !== "fireworks") return;
 
     const payload = event.payload as Record<string, unknown>;
     if (!payload || typeof payload !== "object") return;
 
     let modified = false;
 
-    const tools = payload.tools;
-    if (Array.isArray(tools)) {
-      payload.tools = tools.map((tool: any) => {
-        if (tool?.function?.parameters) {
-          return {
-            ...tool,
-            function: {
-              ...tool.function,
-              parameters: sanitizeSchemaForKimi(tool.function.parameters),
-            },
-          };
-        }
-        return tool;
-      });
+    // Service tier: inject `service_tier` on supported Fireworks models when
+    // priority is selected. Injected at the top level of both the OpenAI
+    // completions and Anthropic Messages request bodies (Fireworks accepts the
+    // field on its chat-completions endpoint; the Anthropic-compatible endpoint
+    // passes it through as a top-level extra).
+    if (currentTier === "priority" && isPriorityApplicable(model.id)) {
+      payload.service_tier = "priority";
       modified = true;
     }
 
-    const responseFormat = payload.response_format as any;
-    if (responseFormat?.json_schema?.schema) {
-      responseFormat.json_schema.schema = sanitizeSchemaForKimi(responseFormat.json_schema.schema);
-      modified = true;
+    // Kimi anchor-bleed sanitization (Kimi K2.x pattern bug). Only applies to
+    // Kimi models, but a single request can be both Kimi and priority-tiered.
+    if (isFireworksKimiModel(model)) {
+      const tools = payload.tools;
+      if (Array.isArray(tools)) {
+        payload.tools = tools.map((tool: any) => {
+          // OpenAI completions shape: tools[].function.parameters
+          if (tool?.function?.parameters) {
+            return {
+              ...tool,
+              function: { ...tool.function, parameters: sanitizeSchemaForKimi(tool.function.parameters) },
+            };
+          }
+          // Anthropic messages shape: tools[].input_schema
+          if (tool?.input_schema) {
+            return { ...tool, input_schema: sanitizeSchemaForKimi(tool.input_schema) };
+          }
+          return tool;
+        });
+        modified = true;
+      }
+
+      // OpenAI-only: the Anthropic Messages API has no response_format equivalent.
+      const responseFormat = payload.response_format as any;
+      if (responseFormat?.json_schema?.schema) {
+        responseFormat.json_schema.schema = sanitizeSchemaForKimi(responseFormat.json_schema.schema);
+        modified = true;
+      }
     }
 
     if (modified) {
@@ -426,4 +622,54 @@ export default function (pi: ExtensionAPI) {
       stripAnchorBleedInPlace(input);
     }
   });
+
+  // Refresh the tier status area when the active model changes.
+  pi.on("model_select", async (_event, ctx) => {
+    updateTierStatus(ctx);
+  });
+
+  // Recompute finalized assistant-message cost against priority pricing so
+  // usage/cost reflects the premium tier rather than the model's base cost.
+  // message_end fires for user, assistant, and toolResult messages; we only
+  // touch assistant messages, and only when priority is active for that model.
+  pi.on("message_end", async (event, _ctx) => {
+    if (currentTier !== "priority") return;
+    const message = (event as any).message;
+    if (!message || message.role !== "assistant") return;
+    const replaced = recomputePriorityCost(message);
+    if (replaced) {
+      return { message: replaced };
+    }
+  });
+
+  // Keybinding: toggle the service tier for the active Fireworks model.
+  pi.registerShortcut(fireworksConfig.serviceTier.keybinding, {
+    description: "Toggle Fireworks service tier (standard / priority)",
+    handler: async (ctx) => {
+      toggleTier(ctx);
+    },
+  });
+
+  // Command form (non-TUI use, or explicit setting): /fireworks-tier [standard|priority|toggle]
+  pi.registerCommand("fireworks-tier", {
+    description: "Set Fireworks service tier: standard | priority | toggle (default)",
+    handler: async (args, ctx) => {
+      const arg = (args || "").trim().toLowerCase();
+      let model: any;
+      try { model = ctx.model; } catch { model = undefined; }
+      if (!model || model.provider !== "fireworks" || !isPriorityApplicable(model.id)) {
+        try { ctx.ui.notify("Fireworks service tier only applies to supported Fireworks models.", "info"); } catch {}
+        return;
+      }
+      if (arg === "priority") {
+        setTier(ctx, "priority");
+        try { ctx.ui.notify("Fireworks service tier: priority", "info"); } catch {}
+      } else if (arg === "standard") {
+        setTier(ctx, "standard");
+        try { ctx.ui.notify("Fireworks service tier: standard", "info"); } catch {}
+      } else {
+        toggleTier(ctx);
+      }
+    },
+  });
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-fireworks-provider",
-  "version": "1.0.2",
+  "version": "1.1.0",
   "description": "Fireworks AI provider extension for pi - Access Kimi, MiniMax, GLM, DeepSeek, and GPT-OSS models through the Fireworks AI API",
   "type": "module",
   "main": "index.ts",

package/patch.json

@@ -1,157 +1,257 @@
 {
+  "accounts/fireworks/models/deepseek-v3p1": {
+    "name": "DeepSeek V3.1",
+    "reasoning": true,
+    "input": ["text"],
+    "cost": {
+      "input": 0.56,
+      "output": 1.68,
+      "cacheRead": 0,
+      "cacheWrite": 0
+    },
+    "contextWindow": 163840,
+    "maxTokens": 163840,
+    "compat": {
+      "supportsReasoningEffort": true
+    }
+  },
+  "accounts/fireworks/models/deepseek-v3p2": {
+    "name": "DeepSeek V3.2",
+    "reasoning": true,
+    "input": ["text"],
+    "cost": {
+      "input": 0.56,
+      "output": 1.68,
+      "cacheRead": 0.28,
+      "cacheWrite": 0
+    },
+    "contextWindow": 163840,
+    "maxTokens": 160000,
+    "compat": {
+      "supportsReasoningEffort": true
+    }
+  },
   "accounts/fireworks/models/deepseek-v4-flash": {
     "name": "DeepSeek V4 Flash",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
       "input": 0.14,
       "output": 0.28,
-      "cacheRead": 0.03,
+      "cacheRead": 0.028,
       "cacheWrite": 0
     },
-    "maxTokens": 1048576,
+    "contextWindow": 1000000,
+    "maxTokens": 384000,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
   "accounts/fireworks/models/deepseek-v4-pro": {
     "name": "DeepSeek V4 Pro",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
-    "maxTokens": 1048576,
+    "input": ["text"],
+    "cost": {
+      "input": 1.74,
+      "output": 3.48,
+      "cacheRead": 0.145,
+      "cacheWrite": 0
+    },
+    "contextWindow": 1000000,
+    "maxTokens": 384000,
+    "compat": {
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
+    }
+  },
+  "accounts/fireworks/models/glm-4p5": {
+    "name": "GLM 4.5",
+    "reasoning": true,
+    "input": ["text"],
+    "cost": {
+      "input": 0.55,
+      "output": 2.19,
+      "cacheRead": 0,
+      "cacheWrite": 0
+    },
+    "contextWindow": 131072,
+    "maxTokens": 131072,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/routers/deepseek-v4-pro": {
-    "name": "DeepSeek V4 Pro (router)",
+  "accounts/fireworks/models/glm-4p5-air": {
+    "name": "GLM 4.5 Air",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 1.74,
-      "output": 3.48,
-      "cacheRead": 0.0145,
+      "input": 0.22,
+      "output": 0.88,
+      "cacheRead": 0,
       "cacheWrite": 0
     },
-    "maxTokens": 1048576,
+    "contextWindow": 131072,
+    "maxTokens": 131072,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/routers/kimi-k2p5-fast": {
-    "name": "Kimi K2.5 Fast (router)",
+  "accounts/fireworks/models/glm-4p7": {
+    "name": "GLM 4.7",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
       "input": 0.6,
-      "output": 3,
-      "cacheRead": 0.1,
+      "output": 2.2,
+      "cacheRead": 0.3,
       "cacheWrite": 0
     },
-    "input": [
-      "text",
-      "image",
-      "video"
-    ],
-    "maxTokens": 256000,
+    "contextWindow": 202752,
+    "maxTokens": 198000,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/routers/kimi-k2p6-turbo": {
-    "name": "Kimi K2.6 Turbo (router)",
+  "accounts/fireworks/models/glm-5": {
+    "name": "GLM 5",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.95,
-      "output": 4,
-      "cacheRead": 0.16,
+      "input": 1,
+      "output": 3.2,
+      "cacheRead": 0.5,
       "cacheWrite": 0
     },
-    "input": [
-      "text",
-      "image"
-    ],
-    "maxTokens": 262144,
+    "contextWindow": 202752,
+    "maxTokens": 131072,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
   "accounts/fireworks/models/glm-5p1": {
     "name": "GLM 5.1",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
       "input": 1.4,
       "output": 4.4,
       "cacheRead": 0.26,
       "cacheWrite": 0
     },
+    "contextWindow": 202800,
     "maxTokens": 131072,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/models/deepseek-v3p2": {
-    "name": "DeepSeek V3.2",
+  "accounts/fireworks/models/glm-5p2": {
+    "name": "GLM 5.2",
+    "api": "openai-completions",
+    "baseUrl": "https://api.fireworks.ai/inference/v1",
     "reasoning": true,
+    "thinkingLevelMap": {
+      "off": "none",
+      "minimal": null,
+      "low": "high",
+      "medium": "high",
+      "xhigh": "max"
+    },
+    "input": ["text"],
     "cost": {
-      "input": 0.56,
-      "output": 1.68,
-      "cacheRead": 0.28,
+      "input": 1.4,
+      "output": 4.4,
+      "cacheRead": 0.26,
       "cacheWrite": 0
     },
-    "maxTokens": 160000,
+    "contextWindow": 1048575,
+    "maxTokens": 131072,
     "compat": {
-      "supportsReasoningEffort": true
+      "supportsStore": false,
+      "supportsDeveloperRole": false
     }
   },
-  "accounts/fireworks/models/minimax-m2p5": {
-    "name": "MiniMax-M2.5",
+  "accounts/fireworks/models/gpt-oss-120b": {
+    "name": "GPT OSS 120B",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.3,
-      "output": 1.2,
-      "cacheRead": 0.03,
+      "input": 0.15,
+      "output": 0.6,
+      "cacheRead": 0.015,
       "cacheWrite": 0
     },
-    "maxTokens": 196608,
+    "contextWindow": 131072,
+    "maxTokens": 32768,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/models/glm-4p5-air": {
-    "name": "GLM 4.5 Air",
+  "accounts/fireworks/models/gpt-oss-20b": {
+    "name": "GPT OSS 20B",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.22,
-      "output": 0.88,
-      "cacheRead": 0,
+      "input": 0.07,
+      "output": 0.3,
+      "cacheRead": 0.035,
       "cacheWrite": 0
     },
-    "maxTokens": 131072,
+    "contextWindow": 131072,
+    "maxTokens": 32768,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/models/glm-5": {
-    "name": "GLM 5",
+  "accounts/fireworks/models/gemma-4-26b-a4b-it": {
+    "name": "Gemma 4 26B A4B IT",
     "reasoning": true,
+    "input": ["text", "image"],
     "cost": {
-      "input": 1,
-      "output": 3.2,
-      "cacheRead": 0.5,
+      "input": 0,
+      "output": 0,
+      "cacheRead": 0,
       "cacheWrite": 0
     },
-    "maxTokens": 131072,
+    "contextWindow": 262000,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/models/deepseek-v3p1": {
-    "name": "DeepSeek V3.1",
+  "accounts/fireworks/models/gemma-4-31b-it": {
+    "name": "Gemma 4 31B IT",
     "reasoning": true,
+    "input": ["text", "image"],
     "cost": {
-      "input": 0.56,
-      "output": 1.68,
+      "input": 0,
+      "output": 0,
       "cacheRead": 0,
       "cacheWrite": 0
     },
-    "maxTokens": 163840,
+    "contextWindow": 262000,
     "compat": {
       "supportsReasoningEffort": true
     }
@@ -159,159 +259,287 @@
   "accounts/fireworks/models/kimi-k2-instruct": {
     "name": "Kimi K2 Instruct",
     "reasoning": false,
+    "input": ["text"],
     "cost": {
       "input": 1,
       "output": 3,
       "cacheRead": 0,
       "cacheWrite": 0
     },
+    "contextWindow": 131072,
     "maxTokens": 16384
   },
-  "accounts/fireworks/models/qwen3p6-plus": {
-    "name": "Qwen 3.6 Plus",
+  "accounts/fireworks/models/kimi-k2-thinking": {
+    "name": "Kimi K2 Thinking",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.5,
+      "input": 0.6,
+      "output": 2.5,
+      "cacheRead": 0.3,
+      "cacheWrite": 0
+    },
+    "contextWindow": 262144,
+    "maxTokens": 256000,
+    "compat": {
+      "supportsReasoningEffort": true
+    }
+  },
+  "accounts/fireworks/models/kimi-k2p5": {
+    "name": "Kimi K2.5",
+    "reasoning": true,
+    "input": ["text", "image"],
+    "cost": {
+      "input": 0.6,
       "output": 3,
       "cacheRead": 0.1,
       "cacheWrite": 0
     },
-    "input": [
-      "text",
-      "image"
-    ],
-    "maxTokens": 8192,
+    "contextWindow": 262144,
+    "maxTokens": 256000,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
+  "accounts/fireworks/models/kimi-k2p6": {
+    "name": "Kimi K2.6",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
+    "reasoning": true,
+    "input": ["text", "image"],
+    "cost": {
+      "input": 0.95,
+      "output": 4,
+      "cacheRead": 0.16,
+      "cacheWrite": 0
+    },
+    "contextWindow": 262000,
+    "maxTokens": 262000,
+    "compat": {
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
+    }
+  },
+  "accounts/fireworks/models/kimi-k2p7-code": {
+    "name": "Kimi K2.7 Code",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
+    "reasoning": true,
+    "input": ["text", "image"],
+    "cost": {
+      "input": 0.95,
+      "output": 4,
+      "cacheRead": 0.19,
+      "cacheWrite": 0
+    },
+    "contextWindow": 262000,
+    "maxTokens": 262000,
+    "compat": {
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
+    }
+  },
   "accounts/fireworks/models/minimax-m2p1": {
     "name": "MiniMax-M2.1",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
       "input": 0.3,
       "output": 1.2,
       "cacheRead": 0.03,
       "cacheWrite": 0
     },
+    "contextWindow": 196608,
     "maxTokens": 200000,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/models/minimax-m2p7": {
-    "name": "MiniMax-M2.7",
+  "accounts/fireworks/models/minimax-m2p5": {
+    "name": "MiniMax-M2.5",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
       "input": 0.3,
       "output": 1.2,
       "cacheRead": 0.03,
       "cacheWrite": 0
     },
+    "contextWindow": 196608,
     "maxTokens": 196608,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/models/glm-4p7": {
-    "name": "GLM 4.7",
+  "accounts/fireworks/models/minimax-m2p7": {
+    "name": "MiniMax-M2.7",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.6,
-      "output": 2.2,
-      "cacheRead": 0.3,
+      "input": 0.3,
+      "output": 1.2,
+      "cacheRead": 0.06,
       "cacheWrite": 0
     },
-    "maxTokens": 198000,
+    "contextWindow": 196608,
+    "maxTokens": 196608,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/models/glm-4p5": {
-    "name": "GLM 4.5",
+  "accounts/fireworks/models/minimax-m3": {
+    "name": "MiniMax-M3",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.55,
-      "output": 2.19,
-      "cacheRead": 0,
+      "input": 0.3,
+      "output": 1.2,
+      "cacheRead": 0.06,
       "cacheWrite": 0
     },
-    "maxTokens": 131072,
+    "contextWindow": 512000,
+    "maxTokens": 512000,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/models/kimi-k2p5": {
-    "name": "Kimi K2.5",
+  "accounts/fireworks/models/qwen3p6-plus": {
+    "name": "Qwen 3.6 Plus",
     "reasoning": true,
+    "input": ["text", "image"],
     "cost": {
-      "input": 0.6,
+      "input": 0.5,
       "output": 3,
       "cacheRead": 0.1,
       "cacheWrite": 0
     },
-    "input": [
-      "text",
-      "image",
-      "video"
-    ],
-    "maxTokens": 256000,
+    "contextWindow": 262144,
+    "maxTokens": 8192,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/models/gpt-oss-20b": {
-    "name": "GPT OSS 20B",
+  "accounts/fireworks/models/qwen3p7-plus": {
+    "name": "Qwen 3.7 Plus",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text", "image"],
     "cost": {
-      "input": 0.05,
-      "output": 0.2,
-      "cacheRead": 0,
+      "input": 0.4,
+      "output": 1.6,
+      "cacheRead": 0.08,
       "cacheWrite": 0
     },
-    "maxTokens": 32768,
+    "contextWindow": 262144,
+    "maxTokens": 65536,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/models/gpt-oss-120b": {
-    "name": "GPT OSS 120B",
+  "accounts/fireworks/routers/deepseek-v4-pro": {
+    "name": "DeepSeek V4 Pro (router)",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.15,
-      "output": 0.6,
-      "cacheRead": 0,
+      "input": 1.74,
+      "output": 3.48,
+      "cacheRead": 0.0145,
       "cacheWrite": 0
     },
-    "maxTokens": 32768,
+    "contextWindow": 1048576,
+    "maxTokens": 1048576,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/models/kimi-k2-thinking": {
-    "name": "Kimi K2 Thinking",
+  "accounts/fireworks/routers/glm-5-fast": {
+    "name": "GLM 5 Fast",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.6,
-      "output": 2.5,
-      "cacheRead": 0.3,
+      "input": 1,
+      "output": 3.2,
+      "cacheRead": 0.5,
       "cacheWrite": 0
     },
-    "maxTokens": 256000,
+    "contextWindow": 202752,
+    "maxTokens": 131072,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/models/kimi-k2p6": {
-    "name": "Kimi K2.6",
+  "accounts/fireworks/routers/glm-5p1-fast": {
+    "name": "GLM 5.1 Fast",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0.95,
-      "output": 4,
-      "cacheRead": 0.16,
+      "input": 2.8,
+      "output": 8.8,
+      "cacheRead": 0.52,
       "cacheWrite": 0
     },
-    "maxTokens": 262144,
+    "contextWindow": 202800,
+    "maxTokens": 131072,
+    "compat": {
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
+    }
+  },
+  "accounts/fireworks/routers/glm-5p2-fast": {
+    "name": "GLM 5.2 Fast",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
+    "reasoning": true,
+    "input": ["text"],
+    "cost": {
+      "input": 2.1,
+      "output": 6.6,
+      "cacheRead": 0.21,
+      "cacheWrite": 0
+    },
+    "contextWindow": 1048575,
+    "maxTokens": 131072,
+    "compat": {
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
+    }
+  },
+  "accounts/fireworks/routers/kimi-k2p5-fast": {
+    "name": "Kimi K2.5 Fast (router)",
+    "reasoning": true,
+    "input": ["text", "image"],
+    "cost": {
+      "input": 0.6,
+      "output": 3,
+      "cacheRead": 0.1,
+      "cacheWrite": 0
+    },
+    "contextWindow": 262144,
+    "maxTokens": 256000,
     "compat": {
       "supportsReasoningEffort": true
     }
@@ -333,84 +561,93 @@
   "accounts/fireworks/routers/kimi-k2p6": {
     "name": "Kimi K2.6 (router)",
     "reasoning": true,
+    "input": ["text", "image"],
     "cost": {
       "input": 0.95,
       "output": 4,
       "cacheRead": 0.16,
       "cacheWrite": 0
     },
-    "input": [
-      "text",
-      "image"
-    ],
+    "contextWindow": 262144,
     "maxTokens": 262144,
     "compat": {
       "supportsReasoningEffort": true
     }
   },
-  "accounts/fireworks/routers/glm-5p1-fast": {
-    "name": "GLM 5.1 Fast (router)",
+  "accounts/fireworks/routers/kimi-k2p6-fast": {
+    "name": "Kimi K2.6 Fast",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text", "image"],
     "cost": {
-      "input": 1.4,
-      "output": 4.4,
-      "cacheRead": 0.26,
-      "cacheWrite": 0
-    },
-    "maxTokens": 131072,
-    "compat": {
-      "supportsReasoningEffort": true
-    }
-  },
-  "accounts/fireworks/routers/minimax-m2p7": {
-    "name": "MiniMax M2.7 (router)",
-    "reasoning": true,
-    "cost": {
-      "input": 0.3,
-      "output": 1.2,
-      "cacheRead": 0.06,
+      "input": 2,
+      "output": 8,
+      "cacheRead": 0.3,
       "cacheWrite": 0
     },
+    "contextWindow": 262000,
+    "maxTokens": 262000,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/routers/glm-5-fast": {
-    "name": "GLM 5 Fast (router)",
+  "accounts/fireworks/routers/kimi-k2p6-turbo": {
+    "name": "Kimi K2.6 Turbo",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text", "image"],
     "cost": {
-      "input": 1,
-      "output": 3.2,
-      "cacheRead": 0.5,
+      "input": 2,
+      "output": 8,
+      "cacheRead": 0.3,
       "cacheWrite": 0
     },
-    "maxTokens": 131072,
+    "contextWindow": 262000,
+    "maxTokens": 262000,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/models/gemma-4-31b-it": {
-    "name": "Gemma 4 31B IT",
+  "accounts/fireworks/routers/kimi-k2p7-code-fast": {
+    "name": "Kimi K2.7 Code Fast",
+    "api": "anthropic-messages",
+    "baseUrl": "https://api.fireworks.ai/inference",
     "reasoning": true,
+    "input": ["text", "image"],
     "cost": {
-      "input": 0,
-      "output": 0,
-      "cacheRead": 0,
+      "input": 1.9,
+      "output": 8,
+      "cacheRead": 0.38,
       "cacheWrite": 0
     },
+    "contextWindow": 262000,
+    "maxTokens": 262000,
     "compat": {
-      "supportsReasoningEffort": true
+      "sendSessionAffinityHeaders": true,
+      "supportsEagerToolInputStreaming": false,
+      "supportsCacheControlOnTools": false,
+      "supportsLongCacheRetention": false
     }
   },
-  "accounts/fireworks/models/gemma-4-26b-a4b-it": {
-    "name": "Gemma 4 26B A4B IT",
+  "accounts/fireworks/routers/minimax-m2p7": {
+    "name": "MiniMax M2.7 (router)",
     "reasoning": true,
+    "input": ["text"],
     "cost": {
-      "input": 0,
-      "output": 0,
-      "cacheRead": 0,
+      "input": 0.3,
+      "output": 1.2,
+      "cacheRead": 0.06,
       "cacheWrite": 0
     },
+    "contextWindow": 204000,
     "compat": {
       "supportsReasoningEffort": true
     }