omnigateway 0.1.7 → 0.1.8

package/gateway.js

@@ -5543,6 +5543,54 @@ function parseJson(raw) {
     return {};
   }
 }
+// packages/ir/src/tokens.ts
+var CHARS_PER_TOKEN = 4;
+var IMAGE_TOKENS = 1600;
+var BLOCK_OVERHEAD = 4;
+var MESSAGE_OVERHEAD = 4;
+function fromText(text) {
+  return Math.ceil(text.length / CHARS_PER_TOKEN);
+}
+function blockTokens(block) {
+  switch (block.type) {
+    case "text":
+      return BLOCK_OVERHEAD + fromText(block.text);
+    case "image":
+      return BLOCK_OVERHEAD + IMAGE_TOKENS;
+    case "thinking":
+      return BLOCK_OVERHEAD + fromText(block.text);
+    case "toolUse":
+      return BLOCK_OVERHEAD + fromText(block.name) + fromText(safeJson(block.input));
+    case "toolResult":
+      return BLOCK_OVERHEAD + fromText(block.toolUseId) + fromText(block.content);
+  }
+}
+function messageTokens(message) {
+  let total = MESSAGE_OVERHEAD;
+  for (const block of message.content)
+    total += blockTokens(block);
+  return total;
+}
+function toolTokens(tool) {
+  return BLOCK_OVERHEAD + fromText(tool.name) + fromText(tool.description ?? "") + fromText(safeJson(tool.inputSchema));
+}
+function safeJson(value) {
+  try {
+    return JSON.stringify(value) ?? "";
+  } catch {
+    return "";
+  }
+}
+function estimateInputTokens(request) {
+  let total = 0;
+  for (const block of request.system ?? [])
+    total += blockTokens(block);
+  for (const message of request.messages)
+    total += messageTokens(message);
+  for (const tool of request.tools ?? [])
+    total += toolTokens(tool);
+  return total;
+}
 // packages/ir/src/validate.ts
 function validateRequest(req) {
   const seenToolUseIds = new Set;
@@ -5573,6 +5621,7 @@ function validateRequest(req) {
 }
 // packages/control/src/config.ts
 var MIN_KEY_LENGTH = 16;
+var TRUTHY = new Set(["1", "true", "yes", "on"]);
 var DECIMAL_INTEGER = /^\d+$/;
 function optionalText(value, fallback) {
   return value?.trim() || fallback;
@@ -5592,6 +5641,7 @@ function loadConfig(env) {
   const baseUrl = optionalText(env.OMNI_BASE_URL, derivedBaseUrl).replace(/\/+$/, "") || derivedBaseUrl;
   const staticDir = env.OMNI_STATIC_DIR?.trim();
   const logFile = env.OMNI_LOG_FILE?.trim();
+  const exposeClaudeCodeAliases = TRUTHY.has((env.OMNI_EXPOSE_CLAUDE_CODE_ALIASES ?? "").trim().toLowerCase());
   const rawLogLevel = env.OMNI_LOG_LEVEL?.trim();
   const logLevel = parseLogLevel(rawLogLevel);
   return {
@@ -5603,9 +5653,14 @@ function loadConfig(env) {
     encryptionKey,
     baseUrl,
     staticDir: staticDir === undefined || staticDir.length === 0 ? null : staticDir,
+    exposeClaudeCodeAliases,
     logFile: logFile === undefined || logFile.length === 0 ? null : logFile
   };
 }
+// packages/providers/src/betas.ts
+var CONTEXT_1M_BETA = "context-1m-2025-08-07";
+var CONTEXT_1M_TOKENS = 1e6;
+
 // packages/providers/src/body.ts
 var BODY_ORDER = {
   anthropic: [
@@ -5713,6 +5768,119 @@ function signAnthropicBody(json) {
   return json.replace(needle, `cch=${computeCch(json)};`);
 }
 
+// packages/providers/src/anthropic/models.ts
+var ANTHROPIC_MODELS = {
+  defaultModel: "claude-opus-5",
+  models: [
+    {
+      id: "claude-fable-5",
+      label: "Claude Fable 5",
+      pricing: { input: 10, output: 50, cacheRead: 1, cacheWrite5m: 12.5, cacheWrite1h: 20 },
+      limits: { contextWindow: 1e6, maxOutputTokens: 128000 }
+    },
+    {
+      id: "claude-opus-5",
+      label: "Claude Opus 5",
+      pricing: { input: 5, output: 25, cacheRead: 0.5, cacheWrite5m: 6.25, cacheWrite1h: 10 },
+      limits: { contextWindow: 1e6, maxOutputTokens: 128000 }
+    },
+    {
+      id: "claude-sonnet-5",
+      label: "Claude Sonnet 5",
+      pricing: { input: 3, output: 15, cacheRead: 0.3, cacheWrite5m: 3.75, cacheWrite1h: 6 },
+      limits: { contextWindow: 1e6, maxOutputTokens: 128000 }
+    },
+    {
+      id: "claude-haiku-4-5",
+      label: "Claude Haiku 4.5",
+      pricing: { input: 1, output: 5, cacheRead: 0.1, cacheWrite5m: 1.25, cacheWrite1h: 2 },
+      limits: { contextWindow: 200000, maxOutputTokens: 64000 }
+    }
+  ]
+};
+
+// packages/providers/src/kimi/models.ts
+var KIMI_MODELS = {
+  defaultModel: "k3-256k",
+  models: [
+    {
+      id: "k3-256k",
+      label: "Kimi K3 \u2014 256K",
+      pricing: { input: 3, output: 15, cacheRead: 0.3, cacheWrite5m: 0, cacheWrite1h: 0 },
+      limits: { contextWindow: 262144, maxOutputTokens: 131072 }
+    },
+    {
+      id: "k3",
+      label: "Kimi K3 \u2014 up to 1M",
+      pricing: { input: 3, output: 15, cacheRead: 0.3, cacheWrite5m: 0, cacheWrite1h: 0 },
+      limits: { contextWindow: 1048576, maxOutputTokens: 131072 }
+    },
+    {
+      id: "kimi-for-coding",
+      label: "Kimi K2.7 Code",
+      pricing: { input: 0.95, output: 4, cacheRead: 0.19, cacheWrite5m: 0, cacheWrite1h: 0 },
+      limits: { contextWindow: 262144, maxOutputTokens: 131072 }
+    },
+    {
+      id: "kimi-for-coding-highspeed",
+      label: "Kimi K2.7 Code \u2014 High Speed",
+      pricing: { input: 0.95, output: 8, cacheRead: 0.19, cacheWrite5m: 0, cacheWrite1h: 0 },
+      limits: { contextWindow: 262144, maxOutputTokens: 131072 }
+    }
+  ]
+};
+
+// packages/providers/src/openai/models.ts
+var OPENAI_MODELS = {
+  defaultModel: "gpt-5.6",
+  models: [
+    {
+      id: "gpt-5.6",
+      label: "GPT-5.6 \u2014 routes to Sol",
+      pricing: { input: 5, output: 30, cacheRead: 0.5, cacheWrite5m: 0, cacheWrite1h: 0 },
+      limits: { contextWindow: 922000, maxOutputTokens: 128000 },
+      oauthLimits: { contextWindow: 272000, maxOutputTokens: 128000 }
+    },
+    {
+      id: "gpt-5.6-sol",
+      label: "GPT-5.6 Sol \u2014 deepest reasoning",
+      pricing: { input: 5, output: 30, cacheRead: 0.5, cacheWrite5m: 0, cacheWrite1h: 0 },
+      limits: { contextWindow: 922000, maxOutputTokens: 128000 },
+      oauthLimits: { contextWindow: 272000, maxOutputTokens: 128000 }
+    },
+    {
+      id: "gpt-5.6-terra",
+      label: "GPT-5.6 Terra \u2014 balanced",
+      pricing: { input: 2, output: 12, cacheRead: 0.2, cacheWrite5m: 0, cacheWrite1h: 0 },
+      limits: { contextWindow: 922000, maxOutputTokens: 128000 },
+      oauthLimits: { contextWindow: 272000, maxOutputTokens: 128000 }
+    },
+    {
+      id: "gpt-5.6-luna",
+      label: "GPT-5.6 Luna \u2014 fastest",
+      pricing: { input: 0.2, output: 1.2, cacheRead: 0.02, cacheWrite5m: 0, cacheWrite1h: 0 },
+      limits: { contextWindow: 922000, maxOutputTokens: 128000 },
+      oauthLimits: { contextWindow: 272000, maxOutputTokens: 128000 }
+    }
+  ]
+};
+
+// packages/providers/src/catalog.ts
+var PROVIDER_MODEL_CATALOG = {
+  anthropic: ANTHROPIC_MODELS,
+  openai: OPENAI_MODELS,
+  kimi: KIMI_MODELS
+};
+function catalogPricing(provider, model) {
+  return PROVIDER_MODEL_CATALOG[provider]?.models.find((entry) => entry.id === model)?.pricing ?? null;
+}
+function catalogLimits(provider, model, auth = "apiKey") {
+  const entry = PROVIDER_MODEL_CATALOG[provider]?.models.find((choice) => choice.id === model);
+  if (entry === undefined)
+    return null;
+  return (auth === "oauth" ? entry.oauthLimits : undefined) ?? entry.limits;
+}
+
 // packages/providers/src/http.ts
 function codeForStatus(status) {
   if (status === 401 || status === 403)
@@ -6270,6 +6438,15 @@ var anthropicAdapter = {
       ["Accept", "text/event-stream"]
     ];
     const betas = new Set(req.request.betas ?? []);
+    const notes = [...degradations];
+    if (betas.has(CONTEXT_1M_BETA)) {
+      const limits = catalogLimits("anthropic", req.model, oauth ? "oauth" : "apiKey");
+      const window2 = limits?.contextWindow;
+      if (window2 !== undefined && window2 < CONTEXT_1M_TOKENS) {
+        betas.delete(CONTEXT_1M_BETA);
+        notes.push("anthropic:context-1m-dropped");
+      }
+    }
     if (oauth) {
       protocol.push(["Authorization", `Bearer ${req.credentials.accessToken}`]);
       betas.add(OAUTH_BETA);
@@ -6297,121 +6474,9 @@ var anthropicAdapter = {
       throw await httpError(res, "anthropic");
     if (res.body === null)
       throw new GatewayError("UPSTREAM", "empty response body", { provider: "anthropic" });
-    return { events: decodeAnthropic(parseSse(res.body)), degradations };
+    return { events: decodeAnthropic(parseSse(res.body)), degradations: notes };
   }
 };
-// packages/providers/src/anthropic/models.ts
-var ANTHROPIC_MODELS = {
-  defaultModel: "claude-opus-5",
-  models: [
-    {
-      id: "claude-fable-5",
-      label: "Claude Fable 5",
-      pricing: { input: 10, output: 50, cacheRead: 1, cacheWrite5m: 12.5, cacheWrite1h: 20 },
-      limits: { contextWindow: 1e6, maxOutputTokens: 128000 }
-    },
-    {
-      id: "claude-opus-5",
-      label: "Claude Opus 5",
-      pricing: { input: 5, output: 25, cacheRead: 0.5, cacheWrite5m: 6.25, cacheWrite1h: 10 },
-      limits: { contextWindow: 1e6, maxOutputTokens: 128000 }
-    },
-    {
-      id: "claude-sonnet-5",
-      label: "Claude Sonnet 5",
-      pricing: { input: 3, output: 15, cacheRead: 0.3, cacheWrite5m: 3.75, cacheWrite1h: 6 },
-      limits: { contextWindow: 1e6, maxOutputTokens: 128000 }
-    },
-    {
-      id: "claude-haiku-4-5",
-      label: "Claude Haiku 4.5",
-      pricing: { input: 1, output: 5, cacheRead: 0.1, cacheWrite5m: 1.25, cacheWrite1h: 2 },
-      limits: { contextWindow: 200000, maxOutputTokens: 64000 }
-    }
-  ]
-};
-
-// packages/providers/src/kimi/models.ts
-var KIMI_MODELS = {
-  defaultModel: "k3-256k",
-  models: [
-    {
-      id: "k3-256k",
-      label: "Kimi K3 \u2014 256K",
-      pricing: { input: 3, output: 15, cacheRead: 0.3, cacheWrite5m: 0, cacheWrite1h: 0 },
-      limits: { contextWindow: 262144, maxOutputTokens: 131072 }
-    },
-    {
-      id: "k3",
-      label: "Kimi K3 \u2014 up to 1M",
-      pricing: { input: 3, output: 15, cacheRead: 0.3, cacheWrite5m: 0, cacheWrite1h: 0 },
-      limits: { contextWindow: 1048576, maxOutputTokens: 131072 }
-    },
-    {
-      id: "kimi-for-coding",
-      label: "Kimi K2.7 Code",
-      pricing: { input: 0.95, output: 4, cacheRead: 0.19, cacheWrite5m: 0, cacheWrite1h: 0 },
-      limits: { contextWindow: 262144, maxOutputTokens: 131072 }
-    },
-    {
-      id: "kimi-for-coding-highspeed",
-      label: "Kimi K2.7 Code \u2014 High Speed",
-      pricing: { input: 0.95, output: 8, cacheRead: 0.19, cacheWrite5m: 0, cacheWrite1h: 0 },
-      limits: { contextWindow: 262144, maxOutputTokens: 131072 }
-    }
-  ]
-};
-
-// packages/providers/src/openai/models.ts
-var OPENAI_MODELS = {
-  defaultModel: "gpt-5.6",
-  models: [
-    {
-      id: "gpt-5.6",
-      label: "GPT-5.6 \u2014 routes to Sol",
-      pricing: { input: 5, output: 30, cacheRead: 0.5, cacheWrite5m: 0, cacheWrite1h: 0 },
-      limits: { contextWindow: 922000, maxOutputTokens: 128000 },
-      oauthLimits: { contextWindow: 272000, maxOutputTokens: 128000 }
-    },
-    {
-      id: "gpt-5.6-sol",
-      label: "GPT-5.6 Sol \u2014 deepest reasoning",
-      pricing: { input: 5, output: 30, cacheRead: 0.5, cacheWrite5m: 0, cacheWrite1h: 0 },
-      limits: { contextWindow: 922000, maxOutputTokens: 128000 },
-      oauthLimits: { contextWindow: 272000, maxOutputTokens: 128000 }
-    },
-    {
-      id: "gpt-5.6-terra",
-      label: "GPT-5.6 Terra \u2014 balanced",
-      pricing: { input: 2, output: 12, cacheRead: 0.2, cacheWrite5m: 0, cacheWrite1h: 0 },
-      limits: { contextWindow: 922000, maxOutputTokens: 128000 },
-      oauthLimits: { contextWindow: 272000, maxOutputTokens: 128000 }
-    },
-    {
-      id: "gpt-5.6-luna",
-      label: "GPT-5.6 Luna \u2014 fastest",
-      pricing: { input: 0.2, output: 1.2, cacheRead: 0.02, cacheWrite5m: 0, cacheWrite1h: 0 },
-      limits: { contextWindow: 922000, maxOutputTokens: 128000 },
-      oauthLimits: { contextWindow: 272000, maxOutputTokens: 128000 }
-    }
-  ]
-};
-
-// packages/providers/src/catalog.ts
-var PROVIDER_MODEL_CATALOG = {
-  anthropic: ANTHROPIC_MODELS,
-  openai: OPENAI_MODELS,
-  kimi: KIMI_MODELS
-};
-function catalogPricing(provider, model) {
-  return PROVIDER_MODEL_CATALOG[provider]?.models.find((entry) => entry.id === model)?.pricing ?? null;
-}
-function catalogLimits(provider, model, auth = "apiKey") {
-  const entry = PROVIDER_MODEL_CATALOG[provider]?.models.find((choice) => choice.id === model);
-  if (entry === undefined)
-    return null;
-  return (auth === "oauth" ? entry.oauthLimits : undefined) ?? entry.limits;
-}
 // packages/providers/src/http-client.ts
 import { request as httpRequest } from "http";
 import { request as httpsRequest } from "https";
@@ -6637,6 +6702,8 @@ function toChatWire(req, model) {
     if (!degradations.includes(d))
       degradations.push(d);
   };
+  if (req.betas?.includes(CONTEXT_1M_BETA))
+    note("kimi:context-1m-dropped");
   const messages = [];
   const system = req.system?.flatMap((b) => b.type === "text" ? [b.text] : []).join(`
 
@@ -6913,6 +6980,8 @@ function toResponsesWire(req, model, opts = { oauth: false }) {
     if (!degradations.includes(d))
       degradations.push(d);
   };
+  if (req.betas?.includes(CONTEXT_1M_BETA))
+    note("openai:context-1m-dropped");
   for (const message of req.messages) {
     const parts = [];
     const inlined = message.role === "system";
@@ -7607,12 +7676,10 @@ async function readSource(deps, source, lines) {
     return "";
   if (source.kind === "file")
     return deps.readFile(source.path, lines) ?? "";
-  const scope = source.scope === "system" ? [] : ["--user"];
+  const unit = source.scope === "system" ? ["-u", source.unit] : [`--user-unit=${source.unit}`];
   const result = await deps.run([
     "journalctl",
-    ...scope,
-    "-u",
-    source.unit,
+    ...unit,
     "-n",
     String(lines),
     "--no-pager",
@@ -21919,7 +21986,9 @@ var dryRunSchema = exports_external.object({
   reasoning: exports_external.boolean().default(false)
 }).strict();
 var modelSchema = exports_external.object({
-  id: exports_external.string().min(1),
+  id: exports_external.string().min(1).refine((value) => !value.toLowerCase().startsWith("claude/"), {
+    message: 'model id must not start with "claude/": that prefix is reserved for discovery mirrors'
+  }),
   strategy: exports_external.enum(["score", "priority", "roundRobin", "weighted"]),
   isAlias: exports_external.boolean(),
   targets: exports_external.array(exports_external.object({
@@ -23373,6 +23442,60 @@ async function createKey(store, input) {
 async function revokeKey(store, id) {
   await store.keys.revoke(id);
 }
+// packages/control/src/modelLimits.ts
+function narrower(a, b) {
+  const context = [a.contextWindow, b.contextWindow].filter((n) => n !== undefined);
+  const output = [a.maxOutputTokens, b.maxOutputTokens].filter((n) => n !== undefined);
+  return {
+    ...context.length === 0 ? {} : { contextWindow: Math.min(...context) },
+    ...output.length === 0 ? {} : { maxOutputTokens: Math.min(...output) }
+  };
+}
+function targetLimits(target, auths) {
+  const ways = auths.size === 0 ? ["apiKey"] : [...auths];
+  let listed = {};
+  for (const auth of ways) {
+    const entry = catalogLimits(target.provider, target.model, auth);
+    if (entry === null)
+      continue;
+    listed = narrower(listed, {
+      contextWindow: entry.contextWindow,
+      maxOutputTokens: entry.maxOutputTokens
+    });
+  }
+  const contextWindow = target.contextWindow ?? listed.contextWindow;
+  const maxOutputTokens = target.maxOutputTokens ?? listed.maxOutputTokens;
+  return {
+    ...contextWindow === undefined ? {} : { contextWindow },
+    ...maxOutputTokens === undefined ? {} : { maxOutputTokens }
+  };
+}
+function servingAuths(credentials) {
+  const byProvider = new Map;
+  for (const credential of credentials) {
+    if (!credential.enabled)
+      continue;
+    const ways = byProvider.get(credential.provider) ?? new Set;
+    ways.add(credential.authType);
+    byProvider.set(credential.provider, ways);
+  }
+  return byProvider;
+}
+function resolveModelLimits(model, credentials) {
+  const auths = servingAuths(credentials);
+  let limits = {};
+  for (const target of model.targets) {
+    limits = narrower(limits, targetLimits(target, auths.get(target.provider) ?? new Set));
+  }
+  return limits;
+}
+function modelDisplayName(model) {
+  const only = model.targets.length === 1 ? model.targets[0] : undefined;
+  if (only === undefined)
+    return model.id;
+  const labelled = PROVIDER_MODEL_CATALOG[only.provider]?.models.find((choice) => choice.id === only.model);
+  return labelled?.label ?? model.id;
+}
 // packages/control/src/models.ts
 async function listModels(store) {
   return store.config.listModels();
@@ -23789,6 +23912,75 @@ async function putSettings(store, input) {
   await store.config.putSettings(settings);
   return settings;
 }
+// packages/control/src/setup.ts
+var KEY_PLACEHOLDER = "<your OmniGateway key>";
+async function describeModelsForSetup(store) {
+  const models = await listModels(store);
+  const credentials = (await listCredentials(store)).map((credential) => ({
+    provider: credential.provider,
+    authType: credential.authType,
+    enabled: credential.enabled
+  }));
+  return models.map((model) => ({
+    model,
+    limits: resolveModelLimits(model, credentials),
+    label: modelDisplayName(model)
+  }));
+}
+function slug(id) {
+  return encodeURIComponent(id).replace(/\./g, "%2E");
+}
+function claudeProfiles(described, input) {
+  const apiKey = input.apiKey ?? KEY_PLACEHOLDER;
+  return described.map(({ model, limits }) => {
+    const useMirror = input.discoveryMirrors === true && !/^(?:claude|anthropic)/i.test(model.id);
+    const modelId = useMirror ? `claude/${model.id}` : model.id;
+    const env2 = {
+      ANTHROPIC_BASE_URL: input.baseUrl,
+      ANTHROPIC_AUTH_TOKEN: apiKey,
+      ANTHROPIC_MODEL: modelId,
+      CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY: "1"
+    };
+    if (limits.contextWindow !== undefined && !/^claude-/i.test(modelId)) {
+      env2.CLAUDE_CODE_MAX_CONTEXT_TOKENS = String(limits.contextWindow);
+    }
+    return {
+      path: `${slug(model.id)}/settings.json`,
+      contents: `${JSON.stringify({ env: env2 }, null, 2)}
+`
+    };
+  });
+}
+function opencodeConfig(described, input) {
+  const models = {};
+  for (const { model, limits, label: label2 } of described) {
+    const limit = limits.contextWindow === undefined ? undefined : {
+      context: limits.contextWindow,
+      ...limits.maxOutputTokens === undefined ? {} : { output: limits.maxOutputTokens }
+    };
+    models[model.id] = { name: label2, ...limit === undefined ? {} : { limit } };
+  }
+  const contents = `${JSON.stringify({
+    $schema: "https://opencode.ai/config.json",
+    provider: {
+      omnigateway: {
+        npm: "@ai-sdk/openai-compatible",
+        name: "OmniGateway",
+        options: {
+          baseURL: `${input.baseUrl.replace(/\/+$/, "").replace(/(?:\/v1)+$/i, "")}/v1`,
+          apiKey: input.apiKey ?? KEY_PLACEHOLDER
+        },
+        models
+      }
+    }
+  }, null, 2)}
+`;
+  return { path: "opencode.json", contents };
+}
+async function setupFiles(store, client, input) {
+  const described = await describeModelsForSetup(store);
+  return client === "claude" ? claudeProfiles(described, input) : [opencodeConfig(described, input)];
+}
 // packages/control/src/tail.ts
 import { closeSync, fstatSync, openSync, readSync, statSync } from "fs";
 var CHUNK = 64 * 1024;
@@ -38503,6 +38695,16 @@ function adminRoutes(deps) {
     await requireAdmin(request2, deps.admin);
     await removeModel(deps.store, params.id);
     return { ok: true };
+  }).get("/api/agent-setup", async ({ request: request2, query }) => {
+    await requireAdmin(request2, deps.admin);
+    const client = query.client === "opencode" ? "opencode" : "claude";
+    return {
+      client,
+      files: await setupFiles(deps.store, client, {
+        baseUrl: deps.baseUrl,
+        discoveryMirrors: deps.discoveryMirrors === true
+      })
+    };
   }).get("/api/keys", async ({ request: request2 }) => {
     await requireAdmin(request2, deps.admin);
     return { keys: await listKeys(deps.store) };
@@ -39361,6 +39563,30 @@ function openaiErrorBody(code, message) {
   return { error: { message, type: e.type, code: e.code } };
 }
 
+// apps/gateway/src/ingress/model.ts
+var DISCOVERY_PREFIX = "claude/";
+var ONE_M_SUFFIX = "[1m]";
+function normalizeClientModel(raw, betas2 = []) {
+  let model = raw.trim();
+  let wantsOneM = false;
+  if (model.toLowerCase().endsWith(ONE_M_SUFFIX)) {
+    const stripped = model.slice(0, -ONE_M_SUFFIX.length).trim();
+    if (stripped.length > 0) {
+      model = stripped;
+      wantsOneM = true;
+    }
+  }
+  if (model.toLowerCase().startsWith(DISCOVERY_PREFIX)) {
+    const stripped = model.slice(DISCOVERY_PREFIX.length);
+    if (stripped.length > 0)
+      model = stripped;
+  }
+  const merged = [...betas2];
+  if (wantsOneM && !merged.includes(CONTEXT_1M_BETA))
+    merged.push(CONTEXT_1M_BETA);
+  return { model, betas: merged };
+}
+
 // apps/gateway/src/ingress/anthropic.ts
 var textBlock = exports_external.object({
   type: exports_external.literal("text"),
@@ -39542,8 +39768,9 @@ function parseAnthropicRequest(body2, headers) {
     text: b.text,
     ...irCacheControl(b.cache_control)
   }));
+  const named = normalizeClientModel(parsed.model, readBetas(headers));
   const request2 = {
-    model: parsed.model,
+    model: named.model,
     messages,
     stream: parsed.stream ?? false
   };
@@ -39574,9 +39801,8 @@ function parseAnthropicRequest(body2, headers) {
   const extras = extraFields(body2, KNOWN);
   if (extras !== undefined)
     request2.vendor = { anthropic: extras };
-  const betas = readBetas(headers);
-  if (betas.length > 0)
-    request2.betas = betas;
+  if (named.betas.length > 0)
+    request2.betas = named.betas;
   return validateRequest(request2);
 }
 
@@ -39725,7 +39951,7 @@ function parseOpenAIRequest(body2) {
     throw new GatewayError("BAD_REQUEST", "messages: at least one non-system message is required");
   }
   const request2 = {
-    model: parsed.model,
+    model: normalizeClientModel(parsed.model).model,
     messages,
     stream: parsed.stream ?? false
   };
@@ -39759,65 +39985,13 @@ function parseOpenAIRequest(body2) {
 
 // apps/gateway/src/routes/models.ts
 var CREATED_AT = new Date(0).toISOString();
-function narrower(a, b) {
-  const context = [a.contextWindow, b.contextWindow].filter((n) => n !== undefined);
-  const output = [a.maxOutputTokens, b.maxOutputTokens].filter((n) => n !== undefined);
-  return {
-    ...context.length === 0 ? {} : { contextWindow: Math.min(...context) },
-    ...output.length === 0 ? {} : { maxOutputTokens: Math.min(...output) }
-  };
-}
-function targetLimits(target, auths) {
-  const ways = auths.size === 0 ? ["apiKey"] : [...auths];
-  let listed = {};
-  for (const auth of ways) {
-    const entry = catalogLimits(target.provider, target.model, auth);
-    if (entry === null)
-      continue;
-    listed = narrower(listed, {
-      contextWindow: entry.contextWindow,
-      maxOutputTokens: entry.maxOutputTokens
-    });
-  }
-  const contextWindow = target.contextWindow ?? listed.contextWindow;
-  const maxOutputTokens = target.maxOutputTokens ?? listed.maxOutputTokens;
-  return {
-    ...contextWindow === undefined ? {} : { contextWindow },
-    ...maxOutputTokens === undefined ? {} : { maxOutputTokens }
-  };
-}
-function limitsOf(model, authsByProvider) {
-  let limits = {};
-  for (const target of model.targets) {
-    limits = narrower(limits, targetLimits(target, authsByProvider.get(target.provider) ?? new Set));
-  }
-  return limits;
-}
-function displayName(model) {
-  const only = model.targets.length === 1 ? model.targets[0] : undefined;
-  if (only === undefined)
-    return model.id;
-  const labelled = PROVIDER_MODEL_CATALOG[only.provider]?.models.find((choice) => choice.id === only.model);
-  return labelled?.label ?? model.id;
-}
-function servingAuths(credentials) {
-  const byProvider = new Map;
-  for (const credential of credentials) {
-    if (!credential.enabled)
-      continue;
-    const ways = byProvider.get(credential.provider) ?? new Set;
-    ways.add(credential.authType);
-    byProvider.set(credential.provider, ways);
-  }
-  return byProvider;
-}
 function describeModel(model, credentials) {
-  const limits = limitsOf(model, servingAuths(credentials));
+  const limits = resolveModelLimits(model, credentials);
   return {
     id: model.id,
     object: "model",
     type: "model",
-    display_name: displayName(model),
+    display_name: modelDisplayName(model),
     created: 0,
     created_at: CREATED_AT,
     owned_by: "omnigateway",
@@ -39825,8 +39999,29 @@ function describeModel(model, credentials) {
     ...limits.maxOutputTokens === undefined ? {} : { max_tokens: limits.maxOutputTokens }
   };
 }
-function modelListBody(models, credentials) {
-  const data = models.map((model) => describeModel(model, credentials));
+var ALREADY_CLAUDE = /^(?:claude|anthropic)/i;
+var MIRROR_PREFIX = "claude/";
+function discoveryMirrors(data) {
+  const taken = new Set(data.map((entry) => entry.id));
+  const mirrors = [];
+  for (const entry of data) {
+    if (ALREADY_CLAUDE.test(entry.id))
+      continue;
+    const id = `${MIRROR_PREFIX}${entry.id}`;
+    if (taken.has(id))
+      continue;
+    mirrors.push({
+      ...entry,
+      id,
+      root: entry.id,
+      display_name: `${entry.display_name} (OmniGateway)`
+    });
+  }
+  return mirrors;
+}
+function modelListBody(models, credentials, options = {}) {
+  const described = models.map((model) => describeModel(model, credentials));
+  const data = options.discoveryMirrors === true ? [...described, ...discoveryMirrors(described)] : described;
   return {
     object: "list",
     data,
@@ -40028,7 +40223,9 @@ function proxyRoutes(deps) {
       const snapshot = await dispatchDeps.snapshots.get(deps.now());
       const models = [...snapshot.models.values()];
       const visibleModels = key.modelAllowlist === null ? models : models.filter((model) => key.modelAllowlist?.includes(model.id));
-      return Response.json(modelListBody(visibleModels, snapshot.credentials));
+      return Response.json(modelListBody(visibleModels, snapshot.credentials, {
+        discoveryMirrors: deps.discoveryMirrors === true
+      }));
     } catch (error51) {
       const gatewayError = asGatewayError(error51);
       logger2.error("model listing failed", {
@@ -40038,6 +40235,25 @@ function proxyRoutes(deps) {
       });
       return errorResponse("anthropic", gatewayError.code, gatewayError.message);
     }
+  }).post("/v1/messages/count_tokens", async ({ request: request2 }) => {
+    try {
+      const key = await authenticateApiKey(deps.store, apiKeyHeader(request2.headers));
+      rateLimiter.consume(key.id, key.rateLimitPerMin);
+      const body2 = await request2.json();
+      const chatRequest = parseAnthropicRequest(body2, request2.headers);
+      if (key.modelAllowlist !== null && !key.modelAllowlist.includes(chatRequest.model)) {
+        throw new GatewayError("AUTH", `model "${chatRequest.model}" is not allowed for this API key`);
+      }
+      return Response.json({ input_tokens: estimateInputTokens(chatRequest) });
+    } catch (error51) {
+      const gatewayError = asGatewayError(error51);
+      logger2.warn("token count failed", {
+        status: HTTP_STATUS[gatewayError.code],
+        code: gatewayError.code,
+        reason: gatewayError.message
+      });
+      return errorResponse("anthropic", gatewayError.code, gatewayError.message);
+    }
   }).onError(({ error: error51 }) => {
     const gatewayError = asGatewayError(error51);
     logger2.error("unhandled proxy route error", {
@@ -40085,10 +40301,13 @@ function createApp(deps) {
     refresh,
     requestId,
     rateLimiter,
-    logger: logger2
+    logger: logger2,
+    discoveryMirrors: deps.discoveryMirrors === true
   })).use(adminRoutes({
     store: deps.store,
     admin,
+    baseUrl: deps.baseUrl,
+    discoveryMirrors: deps.discoveryMirrors === true,
     now,
     sessionTtlMs: ADMIN_SESSION_TTL_MS,
     logger: logger2,
@@ -40331,7 +40550,8 @@ async function main() {
     refresh,
     staticDir,
     logger: logger2,
-    console: console2
+    console: console2,
+    discoveryMirrors: config2.exposeClaudeCodeAliases
   });
   const stopMaintenance = startMaintenance({ store, now, logger: logger2 });
   const stopRefreshScheduler = startRefreshScheduler({ store, refresh, now, logger: logger2 });