Superseded: Add SayGM provider

package.json

@@ -29,6 +29,7 @@
     "kilo:sync": "bun ./packages/core/script/sync-models.ts kilo",
     "llmgateway:sync": "bun ./packages/core/script/sync-models.ts llmgateway",
     "requesty:sync": "bun ./packages/core/script/sync-models.ts requesty",
+    "saygm:sync": "bun ./packages/core/script/sync-models.ts saygm",
     "merge-gateway:sync": "bun ./packages/core/script/sync-models.ts merge-gateway",
     "nano-gpt:sync": "bun ./packages/core/script/sync-models.ts nano-gpt",
     "venice:sync": "bun ./packages/core/script/sync-models.ts venice",

packages/core/src/sync/index.ts

@@ -30,6 +30,7 @@ import { openrouter } from "./providers/openrouter.js";
 import { ovhcloud } from "./providers/ovhcloud.js";
 import { pioneer } from "./providers/pioneer.js";
 import { requesty } from "./providers/requesty.js";
+import { saygm } from "./providers/saygm.js";
 import { tinfoil } from "./providers/tinfoil.js";
 import { vercel } from "./providers/vercel.js";
 import { venice } from "./providers/venice.js";
@@ -139,6 +140,7 @@ export const providers: {
   ovhcloud: SyncProvider<any>;
   pioneer: SyncProvider<any>;
   requesty: SyncProvider<any>;
+  saygm: SyncProvider<any>;
   tinfoil: SyncProvider<any>;
   vercel: SyncProvider<any>;
   venice: SyncProvider<any>;
@@ -170,6 +172,7 @@ export const providers: {
   ovhcloud,
   pioneer,
   requesty,
+  saygm,
   tinfoil,
   vercel,
   venice,
@@ -194,7 +197,7 @@ export const groups = {
     "vercel",
   ],
   cloudflare: ["cloudflare-workers-ai"],
-  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "google", "hyper", "openai", "ovhcloud", "pioneer", "tinfoil", "venice", "wandb", "xai"],
+  direct: ["ambient", "anthropic", "baseten", "chutes", "cortecs", "deepinfra", "digitalocean", "google", "hyper", "openai", "ovhcloud", "pioneer", "saygm", "tinfoil", "venice", "wandb", "xai"],
 } as const;
 
 type ProviderID = keyof typeof providers;

packages/core/src/sync/providers/saygm.ts

@@ -0,0 +1,157 @@
+import { z } from "zod";
+
+import type { ExistingModel, SyncProvider, SyncedModel } from "../index.js";
+
+const API_ENDPOINT = "https://api.saygm.com/v1/models";
+const NDOLLARS_PER_DOLLAR = 1_000_000_000;
+
+const NanoDollars = z.number().int().nonnegative().max(Number.MAX_SAFE_INTEGER);
+
+const PriceDimensions = z.object({
+  input_per_mtok_ndollars: NanoDollars,
+  output_per_mtok_ndollars: NanoDollars,
+  cache_read_per_mtok_ndollars: NanoDollars.optional(),
+  cache_write_5m_per_mtok_ndollars: NanoDollars.optional(),
+  cache_write_1h_per_mtok_ndollars: NanoDollars.optional(),
+  audio_input_per_mtok_ndollars: NanoDollars.optional(),
+  audio_output_per_mtok_ndollars: NanoDollars.optional(),
+  long_context_threshold_tokens: z.number().int().positive().max(Number.MAX_SAFE_INTEGER).optional(),
+  long_context_input_per_mtok_ndollars: NanoDollars.optional(),
+  long_context_output_per_mtok_ndollars: NanoDollars.optional(),
+}).passthrough().superRefine((dimensions, context) => {
+  const longContext = [
+    dimensions.long_context_threshold_tokens,
+    dimensions.long_context_input_per_mtok_ndollars,
+    dimensions.long_context_output_per_mtok_ndollars,
+  ];
+  const present = longContext.filter((value) => value !== undefined).length;
+  if (present !== 0 && present !== longContext.length) {
+    context.addIssue({
+      code: z.ZodIssueCode.custom,
+      message: "Long-context pricing requires a threshold plus input and output prices",
+    });
+  }
+});
+
+const RetailCeiling = z.object({
+  basis: z.literal("published_retail"),
+  dimensions: PriceDimensions,
+}).passthrough();
+
+export const SaygmModel = z.object({
+  id: z.string().min(1),
+  object: z.literal("model"),
+  available: z.boolean(),
+  api_shapes: z.array(z.string().min(1)),
+  price_range: z.object({
+    unit: z.literal("ndollars_per_mtok"),
+    currency: z.literal("USD"),
+    ceiling: RetailCeiling,
+  }).passthrough().optional(),
+}).passthrough();
+
+export const SaygmResponse = z.object({
+  object: z.literal("list"),
+  data: z.array(SaygmModel),
+}).passthrough();
+
+export type SaygmModel = z.infer<typeof SaygmModel>;
+
+export const saygm = {
+  id: "saygm",
+  name: "SayGM",
+  modelsDir: "providers/saygm/models",
+  // The live inventory can change with miner supply. Until the API publishes
+  // durable supply depth, only update models that were reviewed into the
+  // catalog and retain them through temporary outages.
+  skipCreates: true,
+  deleteMissing: false,
+  trackMissingModels: false,
+  sourceID(model) {
+    return model.id;
+  },
+  skippedNotice(ids) {
+    if (ids.length === 0) return [];
+    return [
+      `${ids.length} SayGM models were not created because this sync is update-only until the public API exposes durable miner depth.`,
+      `Skipped remote IDs: ${ids.map((id) => `\`${id}\``).join(", ")}`,
+    ];
+  },
+  missingNotice(paths) {
+    if (paths.length === 0) return [];
+    return [
+      `${paths.length} reviewed SayGM models were retained despite being absent from the current live response.`,
+      `Retained local models: ${paths.map((item) => `\`${item.replace(/\.toml$/, "")}\``).join(", ")}`,
+    ];
+  },
+  async fetchModels() {
+    return fetchSaygmModels();
+  },
+  parseModels(raw) {
+    return SaygmResponse.parse(raw).data;
+  },
+  translateModel(model, context) {
+    if (context.existing(model.id) === undefined) return undefined;
+    const authored = context.authored(model.id);
+    if (authored === undefined) {
+      throw new Error(`SayGM model ${model.id} has no authored catalog entry`);
+    }
+    return {
+      id: model.id,
+      model: buildSaygmModel(model, authored),
+    };
+  },
+} satisfies SyncProvider<SaygmModel>;
+
+export async function fetchSaygmModels(fetcher: typeof fetch = fetch) {
+  const response = await fetcher(API_ENDPOINT);
+  if (!response.ok) {
+    throw new Error(`SayGM models request failed: ${response.status} ${response.statusText}`);
+  }
+  return SaygmResponse.parse(await response.json());
+}
+
+export function buildSaygmModel(model: SaygmModel, existing: ExistingModel): SyncedModel {
+  const ceiling = model.price_range?.ceiling;
+  if (ceiling === undefined) {
+    // A partial catalogue response must not make the hourly update discard or
+    // block a reviewed model. Keep its last reviewed price until SayGM
+    // publishes a complete retail ceiling again.
+    return existing as SyncedModel;
+  }
+
+  const dimensions = ceiling.dimensions;
+  const cost: NonNullable<ExistingModel["cost"]> = {
+    input: usdPerMtok(dimensions.input_per_mtok_ndollars),
+    output: usdPerMtok(dimensions.output_per_mtok_ndollars),
+    cache_read: optionalUsdPerMtok(dimensions.cache_read_per_mtok_ndollars),
+    // models.dev has one cache-write field; its provider catalogs conventionally
+    // use Anthropic's standard five-minute write rate.
+    cache_write: optionalUsdPerMtok(dimensions.cache_write_5m_per_mtok_ndollars),
+    input_audio: optionalUsdPerMtok(dimensions.audio_input_per_mtok_ndollars),
+    output_audio: optionalUsdPerMtok(dimensions.audio_output_per_mtok_ndollars),
+  };
+
+  if (dimensions.long_context_threshold_tokens !== undefined) {
+    const longInput = dimensions.long_context_input_per_mtok_ndollars;
+    const longOutput = dimensions.long_context_output_per_mtok_ndollars;
+    if (longInput === undefined || longOutput === undefined) {
+      throw new Error(`SayGM model ${model.id} has incomplete long-context pricing`);
+    }
+    cost.tiers = [{
+      tier: { type: "context", size: dimensions.long_context_threshold_tokens },
+      input: usdPerMtok(longInput),
+      output: usdPerMtok(longOutput),
+    }];
+  }
+
+  return { ...existing, cost } as SyncedModel;
+}
+
+export function usdPerMtok(ndollars: number) {
+  return ndollars / NDOLLARS_PER_DOLLAR;
+}
+
+function optionalUsdPerMtok(ndollars: number | undefined) {
+  return ndollars === undefined ? undefined : usdPerMtok(ndollars);
+}

packages/core/test/sync.test.ts

@@ -75,6 +75,13 @@ import {
 import { openai, parseOpenAIModels } from "../src/sync/providers/openai.js";
 import { ofox } from "../src/sync/providers/ofox.js";
 import { pioneer } from "../src/sync/providers/pioneer.js";
+import {
+  buildSaygmModel,
+  saygm,
+  SaygmResponse,
+  type SaygmModel,
+  usdPerMtok,
+} from "../src/sync/providers/saygm.js";
 import { google, shouldTrackGoogleModel } from "../src/sync/providers/google.js";
 import { buildTinfoilModel, tinfoil, type TinfoilModel } from "../src/sync/providers/tinfoil.js";
 import { resolveVeniceBaseModel } from "../src/sync/providers/venice.js";
@@ -198,6 +205,109 @@ function readyInceptronModel(overrides: Partial<InceptronModel> = {}): ReadyInce
   })[0]!;
 }
 
+function saygmModel(overrides: Partial<SaygmModel> = {}): SaygmModel {
+  return {
+    id: "gpt-5.4",
+    object: "model",
+    available: true,
+    api_shapes: ["chat.completions", "responses"],
+    price_range: {
+      unit: "ndollars_per_mtok",
+      currency: "USD",
+      ceiling: {
+        basis: "published_retail",
+        dimensions: {
+          input_per_mtok_ndollars: 2_500_000_000,
+          output_per_mtok_ndollars: 15_000_000_000,
+          cache_read_per_mtok_ndollars: 250_000_000,
+          long_context_threshold_tokens: 272_000,
+          long_context_input_per_mtok_ndollars: 5_000_000_000,
+          long_context_output_per_mtok_ndollars: 22_500_000_000,
+        },
+      },
+    },
+    ...overrides,
+  };
+}
+
+test("converts SayGM retail nanodollars into models.dev USD per Mtok costs", () => {
+  const built = buildSaygmModel(saygmModel(), {
+    base_model: "openai/gpt-5.4",
+    reasoning_options: [{ type: "effort", values: ["none", "low", "high"] }],
+  });
+
+  expect(built).toMatchObject({
+    base_model: "openai/gpt-5.4",
+    cost: {
+      input: 2.5,
+      output: 15,
+      cache_read: 0.25,
+      tiers: [{ tier: { type: "context", size: 272_000 }, input: 5, output: 22.5 }],
+    },
+  });
+  expect(usdPerMtok(1)).toBe(0.000000001);
+});
+
+test("uses SayGM's five-minute cache-write rate for models.dev", () => {
+  const built = buildSaygmModel(saygmModel({
+    id: "claude-haiku-4-5",
+    api_shapes: ["messages"],
+    price_range: {
+      unit: "ndollars_per_mtok",
+      currency: "USD",
+      ceiling: {
+        basis: "published_retail",
+        dimensions: {
+          input_per_mtok_ndollars: 1_000_000_000,
+          output_per_mtok_ndollars: 5_000_000_000,
+          cache_read_per_mtok_ndollars: 100_000_000,
+          cache_write_5m_per_mtok_ndollars: 1_250_000_000,
+          cache_write_1h_per_mtok_ndollars: 2_000_000_000,
+        },
+      },
+    },
+  }), { base_model: "anthropic/claude-haiku-4-5" });
+
+  expect(built).toMatchObject({
+    cost: { input: 1, output: 5, cache_read: 0.1, cache_write: 1.25 },
+  });
+});
+
+test("retains reviewed SayGM pricing when the live ceiling is unavailable", () => {
+  const existing = {
+    base_model: "openai/gpt-5.4",
+    cost: { input: 2.5, output: 15, cache_read: 0.25 },
+  };
+
+  const built = buildSaygmModel(saygmModel({ price_range: undefined }), existing);
+
+  expect(built).toEqual(existing);
+});
+
+test("rejects incomplete or non-retail SayGM pricing", () => {
+  const valid = saygmModel();
+  const missingLongOutput = structuredClone(valid);
+  delete missingLongOutput.price_range!.ceiling.dimensions.long_context_output_per_mtok_ndollars;
+  expect(() => SaygmResponse.parse({ object: "list", data: [missingLongOutput] })).toThrow(
+    "Long-context pricing requires a threshold plus input and output prices",
+  );
+
+  const floorOnly = {
+    ...valid,
+    price_range: {
+      ...valid.price_range!,
+      ceiling: { ...valid.price_range!.ceiling, basis: "cheapest_eligible_offer" },
+    },
+  };
+  expect(() => SaygmResponse.parse({ object: "list", data: [floorOnly] })).toThrow();
+});
+
+test("keeps SayGM model lifecycle changes review-only", () => {
+  expect(saygm.skipCreates).toBe(true);
+  expect(saygm.deleteMissing).toBe(false);
+  expect(saygm.trackMissingModels).toBe(false);
+});
+
 test("builds current Inceptron models from explicit base metadata", () => {
   const models = parseInceptronModels({
     object: "list",

providers/saygm/models/claude-fable-5.toml

@@ -0,0 +1,15 @@
+base_model = "anthropic/claude-fable-5"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 10
+output = 50
+cache_read = 1
+cache_write = 12.5
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://api.saygm.com/v1"

providers/saygm/models/claude-haiku-4-5.toml

@@ -0,0 +1,15 @@
+base_model = "anthropic/claude-haiku-4-5"
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1_024
+
+[cost]
+input = 1
+output = 5
+cache_read = 0.1
+cache_write = 1.25
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://api.saygm.com/v1"

providers/saygm/models/claude-opus-4-7.toml

@@ -0,0 +1,15 @@
+base_model = "anthropic/claude-opus-4-7"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://api.saygm.com/v1"

providers/saygm/models/claude-opus-4-8.toml

@@ -0,0 +1,15 @@
+base_model = "anthropic/claude-opus-4-8"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
+
+[provider]
+npm = "@ai-sdk/anthropic"
+api = "https://api.saygm.com/v1"