7.4 billion tokens per month. 34 free LLM providers. 635 free model endpoints. All behind one /v1 endpoint, plus any custom OpenAI-compatible endpoint. Smart routing, automatic failover, encrypted keys. Personal experimentation only.
TypeScript
feat(providers): add IteraCompute free OpenAI-compatible gateway (#1649) (#1655)
tashdroid2359d30bc2d7
diff --git a/client/src/components/keys/shared.tsx b/client/src/components/keys/shared.tsxindex 19870120..9815478a 100644--- a/client/src/components/keys/shared.tsx+++ b/client/src/components/keys/shared.tsx@@ -35,6 +35,7 @@ export const PLATFORMS: { value: Platform; label: string; url: string; keyless?:{ value: 'beatapi', label: 'BeatAPI (JEV decisions; temporary free offer)', url: 'https://beatapi.io' },{ value: 'apinex', label: 'APINex (shared daily free chat and decisions)', url: 'https://apinex.bond/keys' },{ value: 'onomeo', label: 'onomeo (30 free calls / 5 hours, shared)', url: 'https://onomeo.com/dashboard' },+ { value: 'iteracompute', label: 'IteraCompute (30M free tokens / month, shared)', url: 'https://iteracompute.com/free/' },{ value: 'typhoon', label: 'Typhoon (free research API)', url: 'https://playground.opentyphoon.ai' },{ value: 'plugsky', label: 'Plugsky (free chat aliases)', url: 'https://plugsky.com/dashboard' },{ value: 'inferbase', label: 'Inferbase (shared monthly and daily allowance)', url: 'https://inferbase.ai' },diff --git a/client/src/lib/routing.ts b/client/src/lib/routing.tsindex 02f4e983..b3dca104 100644--- a/client/src/lib/routing.ts+++ b/client/src/lib/routing.ts@@ -486,6 +486,7 @@ export const platformColors: Record<string, string> = {beatapi: '#8b5cf6',apinex: '#06b6d4',onomeo: '#f97316',+ iteracompute:'#f59e0b',}// ── Grouped (unified) rendering ──────────────────────────────────────────────diff --git a/docs/ITERACOMPUTE.md b/docs/ITERACOMPUTE.mdnew file mode 100644index 00000000..66937193--- /dev/null+++ b/docs/ITERACOMPUTE.md@@ -0,0 +1,41 @@+# IteraCompute++IteraCompute uses `https://api.iteracompute.com/v1`, with Bearer API-key+authentication, on the OpenAI chat-completions shape. Add a key in the+dashboard or import `ITERACOMPUTE_API_KEY` from an environment file.++- Chat and SSE: `POST /chat/completions`, OpenAI request/response shape.+- Public model list, no key needed: `GET /models` returns every model with+ `is_free` and per-1M-token prices. Credentials are ignored on this endpoint+ (a bogus `Authorization` header still returns `200`, verified live+ 2026-10-22), so it is NOT a credential probe.+- Key validation is off (`supportsKeyValidation === false`): the only+ authenticated endpoint is `/chat/completions`, and every request reserves+ against the monthly allowance before settling (free-tier terms). A health+ probe would burn user quota, so validity is proven by the first served+ request instead; a bogus key fails at request time with `401+ authentication_error invalid_api_key` (verified live 2026-10-22).++## Free eligibility, verified 2026-10-22++The free tier is one account-wide monthly TOKEN allowance: `qwen/qwen3.8-27b`+gets 30M tokens/month (input + output, reasoning included) at 160,000 TPM with+128K free context. No daily quota, no paid overage: once the allowance is spent+the API answers 429 until the reset at 00:00 UTC on the 1st (no rollover).+Free accounts may queue at peak; full or timed-out queues return 429 with+`Retry-After`, infrastructure failures 503. For plain answers without reasoning+text, send `"chat_template_kwargs": {"enable_thinking": false}`.++`moonshotai/kimi-k3-free` also shows `is_free: true` on the model list but+requires a $10 cumulative recharge first, so it does not meet the never-paid+rule from #853 and stays out of the catalog.++Model rows are published only through the signed Oracle catalog. No seeds,+model-discovery enrollment, or database migration bypasses the existing+Premium-now / Free-after-30-days release gate.++Sources:++- <https://iteracompute.com/free/>+- <https://iteracompute.com/free-tier-terms/>+- issue #1649diff --git a/server/src/__tests__/services/iteracompute.test.ts b/server/src/__tests__/services/iteracompute.test.tsnew file mode 100644index 00000000..e2cf93d0--- /dev/null+++ b/server/src/__tests__/services/iteracompute.test.ts@@ -0,0 +1,69 @@+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';+import { initDb, getDb } from '../../db/index.js';+import { getProvider } from '../../providers/index.js';+import { IteraComputeProvider, ITERACOMPUTE_BASE_URL } from '../../providers/iteracompute.js';+import { AUTH_JSON_PROVIDER_MAP, parseKeysFromFile } from '../../lib/key-parser.js';+import { inferPoolForPlatform } from '../../services/provider-quota.js';++const json = (body: unknown, status = 200) => new Response(JSON.stringify(body), { status });+const completion = (model?: string) => ({ id: 'c', object: 'chat.completion', created: 1, model,+ choices: [{ index: 0, message: { role: 'assistant', content: 'OK' }, finish_reason: 'stop' }],+ usage: { prompt_tokens: 3, completion_tokens: 1, total_tokens: 4 } });++beforeEach(() => { process.env.ENCRYPTION_KEY = '0'.repeat(64); initDb(':memory:'); });+afterEach(() => { vi.restoreAllMocks(); getDb().close(); });++describe('iteracompute registration and import (#1649)', () => {+ it('registers the adapter, the env import alias and one shared monthly pool', () => {+ expect(getProvider('iteracompute')).toBeInstanceOf(IteraComputeProvider);+ expect(getProvider('iteracompute')!.keyless).toBe(false);+ expect(parseKeysFromFile('ITERACOMPUTE_API_KEY=test-key', 'keys.env').keys[0].platform).toBe('iteracompute');+ expect(AUTH_JSON_PROVIDER_MAP.iteracompute).toBe('iteracompute');+ // One 30M-token monthly allowance shared across every free model, so a 429+ // on one model benches the platform, not the model.+ expect(inferPoolForPlatform('iteracompute', 'qwen/qwen3.8-27b'))+ .toBe(inferPoolForPlatform('iteracompute', 'qwen/qwen3.9-30b'));+ });+});++describe('iteracompute key validation (#1649)', () => {+ // GET /models is anonymous (a bogus Bearer header still returns 200, verified+ // live 2026-10-22) and /chat/completions reserves monthly allowance per+ // request, so there is no free credential probe — same stance as freeai+ // (#1459): inconclusive, never a false verdict, never quota spent.+ it('never probes: validation is inconclusive by design', async () => {+ const fetch = vi.spyOn(global, 'fetch');+ const provider = new IteraComputeProvider();+ expect(provider.supportsKeyValidation).toBe(false);+ await expect(provider.validateKey('test-key')).rejects.toThrow(/no free credential-validation endpoint/i);+ // The probe must never hit the network: no free credential endpoint exists.+ expect(fetch).not.toHaveBeenCalled();+ });+});++describe('iteracompute chat', () => {+ it('forwards OpenAI-shape completions with attribution', async () => {+ const fetch = vi.spyOn(global, 'fetch').mockResolvedValue(json(completion('qwen/qwen3.8-27b')));+ const result = await new IteraComputeProvider().chatCompletion('test-key', [{ role: 'user', content: 'Hi' }], 'qwen/qwen3.8-27b', { max_tokens: 512 });+ expect(fetch.mock.calls[0][0]).toBe(`${ITERACOMPUTE_BASE_URL}/chat/completions`);+ expect(JSON.parse(String(fetch.mock.calls[0][1]?.body))).toMatchObject({ model: 'qwen/qwen3.8-27b', max_tokens: 512 });+ expect(result.choices[0].message.content).toBe('OK');+ expect(result._routed_via).toEqual({ platform: 'iteracompute', model: 'qwen/qwen3.8-27b' });+ });++ // A bad key is only discovered at request time (validation is off), so the+ // 401 must surface as an invalid-key verdict through the normal error path.+ it('surfaces the live 401 invalid_api_key shape as an upstream error', async () => {+ vi.spyOn(global, 'fetch').mockResolvedValue(json({ error: { message: 'IteraCompute authentication failed', type: 'authentication_error', code: 'invalid_api_key' } }, 401));+ await expect(new IteraComputeProvider().chatCompletion('bad-key', [{ role: 'user', content: 'Hi' }], 'qwen/qwen3.8-27b'))+ .rejects.toMatchObject({ status: 401 });+ });++ // Full/timed-out free queues answer 429 + Retry-After per the free-tier+ // terms; the status must reach the fallback loop intact.+ it('surfaces a 429 queue rejection as a retryable upstream error', async () => {+ vi.spyOn(global, 'fetch').mockResolvedValue(json({ error: { message: 'queue full' } }, 429));+ await expect(new IteraComputeProvider().chatCompletion('test-key', [{ role: 'user', content: 'Hi' }], 'qwen/qwen3.8-27b'))+ .rejects.toMatchObject({ status: 429 });+ });+});diff --git a/server/src/lib/key-parser.ts b/server/src/lib/key-parser.tsindex cf3128a1..03f8f00c 100644--- a/server/src/lib/key-parser.ts+++ b/server/src/lib/key-parser.ts@@ -49,6 +49,7 @@ export const PREFIX_MAP: Record<string, string> = {BEATAPI_: 'beatapi',APINEX_: 'apinex',ONOMEO_: 'onomeo',+ ITERACOMPUTE_: 'iteracompute',BEAT_API_: 'beatapi',SERPSEARCH_: 'serpsearch',SERP_SEARCH_: 'serpsearch',@@ -166,6 +167,7 @@ export const AUTH_JSON_PROVIDER_MAP: Record<string, string> = {beatapi: 'beatapi',apinex: 'apinex',onomeo: 'onomeo',+ iteracompute: 'iteracompute',serpsearch: 'serpsearch',typhoon: 'typhoon',plugsky: 'plugsky',diff --git a/server/src/providers/index.ts b/server/src/providers/index.tsindex 4f992d85..409a30e7 100644--- a/server/src/providers/index.ts+++ b/server/src/providers/index.ts@@ -41,6 +41,7 @@ import { LogfareProvider } from './logfare.js';import { BeatApiProvider } from './beatapi.js';import { ApinexProvider } from './apinex.js';import { OnomeoProvider } from './onomeo.js';+import { IteraComputeProvider } from './iteracompute.js';import { SerpSearchProvider } from './serpsearch.js';const providers = new Map<Platform, BaseProvider>();@@ -88,6 +89,9 @@ register(new MoondreamProvider());register(new BeatApiProvider());register(new ApinexProvider());register(new OnomeoProvider());+// IteraCompute (#1649): free OpenAI-compatible gateway, one account-wide+// monthly token allowance; no free credential probe (see adapter docs).+register(new IteraComputeProvider());register(new SerpSearchProvider());// Free-plan grants are shared wallets, not free credits per model. Eligibilitydiff --git a/server/src/providers/iteracompute.ts b/server/src/providers/iteracompute.tsnew file mode 100644index 00000000..0397466c--- /dev/null+++ b/server/src/providers/iteracompute.ts@@ -0,0 +1,42 @@+import type { KeyValidationResult } from './base.js';+import { OpenAICompatProvider } from './openai-compat.js';++export const ITERACOMPUTE_BASE_URL = 'https://api.iteracompute.com/v1';++/**+ * IteraCompute (#1649): an OpenAI-compatible gateway whose free tier is one+ * account-wide monthly TOKEN allowance — 30M tokens (input + output, reasoning+ * included) per account, 160,000 TPM, resetting at 00:00 UTC on the 1st with+ * no rollover and no paid overage (free-tier terms, Oct 2026). The free model+ * published today is `qwen/qwen3.8-27b` (`is_free: true`, zero prices on the+ * public model list; the paid rows share the key, so the 30-day free gate and+ * the signed catalog decide what actually routes here).+ *+ * Key validation is deliberately OFF (mirrors freeai #1459): `GET /models` is+ * anonymous — a bogus `Authorization` header still returns 200 (verified live+ * 2026-10-22) — and the only authenticated endpoint is `/chat/completions`,+ * which reserves against the monthly allowance on every request (the terms+ * say requests reserve their maximum usage before settling). A health probe+ * would therefore either prove nothing or burn user quota, so validity is+ * established by the first served request instead (markKeyHealthyFromRequest).+ * A real call with a bad key does answer `401 authentication_error+ * invalid_api_key`, which is what fails it over at request time.+ */+export class IteraComputeProvider extends OpenAICompatProvider {+ constructor() {+ // Free capacity is shared and queues at peak times (429 + Retry-After when+ // full), so the first byte can be slower than the 60s compat default.+ super({ platform: 'iteracompute', name: 'IteraCompute', baseUrl: ITERACOMPUTE_BASE_URL, timeoutMs: 120_000 });+ }++ override get supportsKeyValidation(): boolean {+ return false;+ }++ override async validateKey(_apiKey: string): Promise<KeyValidationResult> {+ // No endpoint distinguishes genuine from bogus keys for free: /models+ // ignores credentials, /chat/completions reserves monthly allowance.+ // Throw so the health pass keeps the key's current state unknown.+ throw new Error('IteraCompute has no free credential-validation endpoint; key validity is unknown (no quota spent)');+ }+}diff --git a/server/src/routes/keys.ts b/server/src/routes/keys.tsindex 0e8110b1..aadccb3e 100644--- a/server/src/routes/keys.ts+++ b/server/src/routes/keys.ts@@ -65,6 +65,7 @@ const PLATFORMS = ['beatapi','apinex','onomeo',+ 'iteracompute','serpsearch','freeai','google', 'groq', 'cerebras', 'sail', 'electronhub', 'experiential', 'router9', 'septor', 'clod', 'speechify', 'blaze', 'lucidity', 'airforce', 'dreamprompting', 'waterfall', 'logfare', 'bai', 'radeon', 'nvidia', 'mistral',diff --git a/server/src/services/provider-quota.ts b/server/src/services/provider-quota.tsindex d5df50d5..da9c12c9 100644--- a/server/src/services/provider-quota.ts+++ b/server/src/services/provider-quota.ts@@ -122,6 +122,9 @@ export function inferPoolForPlatform(platform: Platform, modelId?: string | null// onomeo: one 30-calls/5-hour account allowance across every free model,// so one pool (issue #1460).if (platform === 'onomeo') return 'onomeo::five-hour-free';+ // IteraCompute: one 30M-token monthly account allowance across every free+ // model (input + output combined), so one pool (issue #1649).+ if (platform === 'iteracompute') return 'iteracompute::monthly-free';if (platform === 'plugsky') return 'plugsky::fair-use';if (platform === 'inferbase') return 'inferbase::managed-free';if (platform === 'simplellm') return 'simplellm::free';@@ -186,6 +189,7 @@ export function inferPoolForPlatform(platform: Platform, modelId?: string | nullfunction isSharedPool(platform: Platform): boolean {if (platform === 'apinex') return true;if (platform === 'onomeo') return true;+ if (platform === 'iteracompute') return true;if (platform === 'aclide') return true;if (platform === 'speka') return true;if (platform === 'plugsky') return true;diff --git a/shared/types.ts b/shared/types.tsindex 08409bf5..fa2b14f9 100644--- a/shared/types.ts+++ b/shared/types.ts@@ -79,6 +79,9 @@ export type Platform =// OpenAI-compatible gateway; free models share a 30-calls/5-hour account// allowance on top of a site-wide pool (issue #1460).| 'onomeo'+ // OpenAI-compatible gateway; the free tier is one account-wide monthly+ // token allowance (30M, resets monthly; issue #1649).+ | 'iteracompute'// Ongoing free research API for chat and speech recognition; catalog only.| 'typhoon'// Permanent free chat aliases, shared fair-use quota; signed catalog only.