mohdel 3.6.1 → 3.7.0

package/js/chatgpt/index.js

@@ -4,6 +4,7 @@ import { hostname } from 'node:os'
 import { createRemoteJWKSet, jwtVerify } from 'jose'
 import { defaultDirectory, readStore, withStore } from './store.js'
 import { discoverModels, visibleModels } from './models.js'
+import providers from '../../src/lib/providers.js'
 
 const ISSUER = 'https://auth.openai.com'
 const RESOURCE = 'https://api.openai.com/v1'
@@ -15,7 +16,7 @@ const random = () => randomBytes(32).toString('base64url')
 const REFRESH_MARGIN = 30000
 // Every registration made before names were stored was sent this hint.
 const UNNAMED_REGISTRATION = 'Mohdel'
-export const usageURL = 'https://chatgpt.com/settings/usage'
+export const usageURL = providers.chatgpt.billing.usage
 
 async function jsonRequest (fetcher, url, options = {}) {
   let response

package/js/session/adapters/_pricing.js

@@ -10,6 +10,16 @@
  */
 
 import { setCatalog, specFor } from './_catalog.js'
+import { providerOf } from '#core/model-id.js'
+import providers from '../../../src/lib/providers.js'
+
+// `plan` and `capacity` calls have no per-call API invoice, so `cost`, which is
+// API USD, is 0 whatever the catalog prices. `fake` and `echo` are adapters,
+// not providers: they carry no declaration and price from the catalog.
+function unmetered (model) {
+  const def = providers[providerOf(model)]
+  return def !== undefined && def.billing.kind !== 'metered'
+}
 
 /**
  * Pure cost computation from spec + usage.
@@ -112,11 +122,30 @@ function resolveTier (price, tokens) {
  * @returns {number}
  */
 export function costFor (envelope, usage) {
-  // Plan usage has no per-call API invoice; it still consumes the plan allowance.
-  if (envelope.model.startsWith('chatgpt/')) return 0
+  if (unmetered(envelope.model)) return 0
   return computeCost(specFor(envelope), usage)
 }
 
+/**
+ * @param {{model: string}} envelope
+ * @param {any} spec
+ * @param {{durationSeconds?: number | null, inputTokens?: number, outputTokens?: number}} usage
+ * @returns {number}
+ */
+export function transcriptionCostFor (envelope, spec, usage) {
+  return unmetered(envelope.model) ? 0 : computeTranscriptionCost(spec, usage)
+}
+
+/**
+ * @param {{model: string}} envelope
+ * @param {any} spec
+ * @param {{inputTokens?: number}} usage
+ * @returns {number}
+ */
+export function embeddingCostFor (envelope, spec, usage) {
+  return unmetered(envelope.model) ? 0 : computeEmbeddingCost(spec, usage)
+}
+
 /**
  * Cost of a transcription call.
  *

package/js/session/adapters/embedding/cohere.js

@@ -12,7 +12,7 @@
 
 import { getSpec } from '../_catalog.js'
 import { classifyProviderError, fromHttpStatus, typedError } from '../_errors.js'
-import { computeEmbeddingCost } from '../_pricing.js'
+import { embeddingCostFor } from '../_pricing.js'
 import { catalogKey, bareOf } from '#core/model-id.js'
 import { checkBatch, checkDimensions, resolveInputType, widthOf } from './_shared.js'
 
@@ -85,7 +85,7 @@ export async function cohereEmbedding (envelope, deps = {}) {
     dimensions: widthOf(vectors),
     inputType,
     inputTokens,
-    cost: computeEmbeddingCost(spec, { inputTokens }),
+    cost: embeddingCostFor(envelope, spec, { inputTokens }),
     timestamps: { start, first: end, end }
   }
 }

package/js/session/adapters/embedding/gemini.js

@@ -13,7 +13,7 @@
 
 import { getSpec } from '../_catalog.js'
 import { classifyProviderError, fromHttpStatus, typedError } from '../_errors.js'
-import { computeEmbeddingCost } from '../_pricing.js'
+import { embeddingCostFor } from '../_pricing.js'
 import { catalogKey, bareOf } from '#core/model-id.js'
 import { checkBatch, checkDimensions, resolveInputType, widthOf } from './_shared.js'
 
@@ -86,7 +86,7 @@ export async function geminiEmbedding (envelope, deps = {}) {
     dimensions: widthOf(vectors),
     inputType: taskType,
     inputTokens,
-    cost: computeEmbeddingCost(spec, { inputTokens }),
+    cost: embeddingCostFor(envelope, spec, { inputTokens }),
     timestamps: { start, first: end, end }
   }
 }

package/js/session/adapters/embedding/openai_compatible.js

@@ -11,7 +11,7 @@
 
 import { getSpec } from '../_catalog.js'
 import { classifyProviderError, fromHttpStatus, typedError } from '../_errors.js'
-import { computeEmbeddingCost } from '../_pricing.js'
+import { embeddingCostFor } from '../_pricing.js'
 import { catalogKey, bareOf } from '#core/model-id.js'
 import { checkBatch, checkDimensions, resolveInputType, widthOf } from './_shared.js'
 
@@ -85,7 +85,7 @@ export function createEmbeddingAdapter ({ baseURL, dimensionsField = 'dimensions
       dimensions: widthOf(vectors),
       inputType,
       inputTokens,
-      cost: computeEmbeddingCost(spec, { inputTokens }),
+      cost: embeddingCostFor(envelope, spec, { inputTokens }),
       timestamps: { start, first: end, end }
     }
   }

package/js/session/adapters/transcription/openai_compatible.js

@@ -23,7 +23,7 @@ import { basename } from 'node:path'
 import { getSpec } from '../_catalog.js'
 import { classifyProviderError, fromHttpStatus, typedError } from '../_errors.js'
 import { dataUriPayload, isTrustedMedia, mediaScheme, readLocalMedia } from '../_media.js'
-import { computeTranscriptionCost } from '../_pricing.js'
+import { transcriptionCostFor } from '../_pricing.js'
 import { catalogKey, bareOf } from '#core/model-id.js'
 
 /**
@@ -67,7 +67,7 @@ export function createTranscriptionAdapter ({ baseURL, responseFormat }) {
     const body = await res.json()
     const durationSeconds = extractDuration(body)
     const tokens = extractTokens(body)
-    const cost = computeTranscriptionCost(spec, { durationSeconds, ...tokens })
+    const cost = transcriptionCostFor(envelope, spec, { durationSeconds, ...tokens })
 
     const end = String(process.hrtime.bigint())
     return {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mohdel",
-  "version": "3.6.1",
+  "version": "3.7.0",
   "license": "MIT",
   "author": {
     "name": "Christophe Le Bars",
@@ -144,7 +144,7 @@
     "@opentelemetry/exporter-trace-otlp-grpc": "^0.222.0",
     "@opentelemetry/sdk-node": "^0.222.0",
     "chalk": "^6.0.1",
-    "mohdel-thin-gate-linux-x64-gnu": "3.6.1"
+    "mohdel-thin-gate-linux-x64-gnu": "3.7.0"
   },
   "dependencies": {
     "@anthropic-ai/sdk": "^0.129.0",

package/src/cli/ask.js

@@ -1,5 +1,6 @@
 import mohdel, { silent } from '../lib/index.js'
 import { getConfig, loadDefaultEnv } from '../lib/common.js'
+import providerDefs, { billingOf } from '../lib/providers.js'
 
 const noop = () => {}
 
@@ -47,8 +48,10 @@ export const hintsForError = (err, modelId) => {
     else hints.push('→ run:  mo                          # interactive provider/key setup')
   }
 
-  if (provider === 'chatgpt' && /RATE_LIMIT|QUOTA_EXHAUSTED|429|usage limit/i.test(`${err?.type || ''} ${both}`)) {
-    hints.push('→ manage ChatGPT plan usage: https://chatgpt.com/settings/usage')
+  // The id is what the user typed, so its provider may not exist.
+  const usage = providerDefs[provider]?.billing.usage
+  if (usage && /RATE_LIMIT|QUOTA_EXHAUSTED|429|usage limit/i.test(`${err?.type || ''} ${both}`)) {
+    hints.push(`→ manage ${providerDefs[provider].billing.label} usage: ${usage}`)
   }
 
   if (/deprecated/i.test(both) && /replacement/i.test(both)) {
@@ -223,7 +226,8 @@ Examples:
     if (tokens.inputTokens) summary.push(`${tokens.inputTokens} in`)
     if (tokens.outputTokens) summary.push(`${tokens.outputTokens} out`)
     if (tokens.thinkingTokens) summary.push(`${tokens.thinkingTokens} think`)
-    if (model.id.startsWith('chatgpt/')) summary.push('Using ChatGPT plan — manage usage: https://chatgpt.com/settings/usage')
+    const billing = billingOf(model.id)
+    if (billing.kind !== 'metered') summary.push(`not metered · ${billing.label}${billing.usage ? ` — manage usage: ${billing.usage}` : ''}`)
     else if (tokens.cost != null) summary.push(`$${tokens.cost.toFixed(4)}`)
     if (tokens.speed) {
       const served = tokens.servedSpeed

package/src/cli/model.js

@@ -731,11 +731,11 @@ Alibaba's Qwen. Routing follows the provider — see "mo provider --help".`)
 
 // The provider API returns ids, never prices — curated entries land unpriced.
 function printCurateNext (providerName) {
-  if (providerName === 'chatgpt') {
-    console.log('ChatGPT plan models are ready. Use mo ask chatgpt/<model-slug> "your prompt". Plan usage is not API billing.')
+  const def = providerDefs[providerName]
+  if (def.billing.kind !== 'metered') {
+    console.log(`${providerName} models are ready. Use mo ask ${providerName}/<model-id> "your prompt". Calls are not metered: they draw on your ${def.billing.label}.`)
     return
   }
-  const def = providerDefs[providerName]
   if (def?.pricesFromApi) {
     console.log(`\n${meta(`${providerName} publishes prices in its model list — the entries are complete.`)}`)
     console.log(`${meta('Check them:')} mo ls  ${meta('│')}  mo check`)
@@ -779,6 +779,9 @@ const SINGLE_DIMENSION_PRICES = [
 ]
 
 function formatPrice (info) {
+  // An entry naming an unknown provider is `mo check`'s to report; it lists with its prices.
+  const billing = providerDefs[info.provider]?.billing
+  if (billing && billing.kind !== 'metered') return meta(`not metered · ${billing.label}`)
   const inp = resolvePrice(info.inputPrice)
   const out = resolvePrice(info.outputPrice)
   if (inp || out) return price(`$${inp}`) + meta('/') + price(`$${out}`)

package/src/lib/catalog-review.js

@@ -73,6 +73,9 @@ export const reviewEntry = (key, spec, catalog, { strict = false, local = null }
   if (spec.provider && spec.provider !== keyProvider) {
     errors.push(`${key}: spec.provider '${spec.provider}' doesn't match key prefix '${keyProvider}'`)
   }
+  if (providerConfig?.billing.kind === 'capacity' && Object.keys(spec).some(k => k.endsWith('Price'))) {
+    warnings.push(`${key}: prices are never charged — ${providerConfig.billing.label} calls report cost 0`)
+  }
   if (providerConfig && spec.sdk && spec.sdk !== providerConfig.sdk) {
     errors.push(`${key}: spec.sdk '${spec.sdk}' doesn't match provider sdk '${providerConfig.sdk}'`)
   }

package/src/lib/providers.js

@@ -1,3 +1,5 @@
+import { providerOf } from '#core/model-id.js'
+
 const LOCAL_API_KEY_ENV = 'MOHDEL_LOCAL_API_SK'
 
 // `contextSemantics` and `outputCapStrategy` are published facts about a
@@ -6,8 +8,13 @@ const LOCAL_API_KEY_ENV = 'MOHDEL_LOCAL_API_SK'
 // own provider requests does not have to rediscover the behaviour one 400 at a
 // time. Entries may override `outputCapStrategy` per model. See
 // ARCHITECTURE.md > "The output budget is capped to the model's ceiling".
+// `billing` is how a provider's calls are paid for. `metered`: API money,
+// reported as `cost`. `plan`: a share of a subscription allowance, seen at
+// `usage`. `capacity`: hardware run or rented at a flat rate. `cost` is 0 for
+// the last two.
 const providers = {
   anthropic: {
+    billing: { kind: 'metered' },
     sdk: 'anthropic',
     apiKeyEnv: 'ANTHROPIC_API_SK',
     createConfiguration: apiKey => ({ apiKey }),
@@ -20,6 +27,7 @@ const providers = {
     outputCapStrategy: 'error'
   },
   cerebras: {
+    billing: { kind: 'metered' },
     sdk: 'cerebras',
     apiKeyEnv: 'CEREBRAS_API_SK',
     createConfiguration: apiKey => ({ apiKey }),
@@ -32,6 +40,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   chatgpt: {
+    billing: { kind: 'plan', label: 'ChatGPT plan', usage: 'https://chatgpt.com/settings/usage' },
     sdk: 'openai',
     catalogClient: 'chatgpt',
     refreshConfiguration: true,
@@ -47,6 +56,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   deepseek: {
+    billing: { kind: 'metered' },
     sdk: 'openai',
     api: 'chatCompletions',
     apiKeyEnv: 'DEEPSEEK_API_SK',
@@ -61,6 +71,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   fireworks: {
+    billing: { kind: 'metered' },
     sdk: 'fireworks',
     apiKeyEnv: 'FIREWORKS_API_SK',
     baseURL: 'https://api.fireworks.ai/inference/v1',
@@ -74,6 +85,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   gemini: {
+    billing: { kind: 'metered' },
     sdk: 'gemini',
     apiKeyEnv: 'GEMINI_API_SK',
     createConfiguration: apiKey => ({ apiKey }),
@@ -86,6 +98,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   groq: {
+    billing: { kind: 'metered' },
     sdk: 'groq',
     apiKeyEnv: 'GROQ_API_SK',
     createConfiguration: apiKey => ({ apiKey }),
@@ -96,6 +109,7 @@ const providers = {
     }
   },
   local: {
+    billing: { kind: 'capacity', label: 'local server' },
     sdk: 'openai',
     api: 'chatCompletions',
     catalog: false,
@@ -105,6 +119,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   meta: {
+    billing: { kind: 'metered' },
     sdk: 'openai',
     apiKeyEnv: 'META_API_SK',
     baseURL: 'https://api.meta.ai/v1',
@@ -117,6 +132,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   cohere: {
+    billing: { kind: 'metered' },
     sdk: 'cohere',
     api: 'embeddings',
     apiKeyEnv: 'COHERE_API_SK',
@@ -129,6 +145,7 @@ const providers = {
     }
   },
   mistral: {
+    billing: { kind: 'metered' },
     sdk: 'openai',
     api: 'chatCompletions',
     apiKeyEnv: 'MISTRAL_API_SK',
@@ -141,6 +158,7 @@ const providers = {
     }
   },
   novita: {
+    billing: { kind: 'metered' },
     sdk: 'openai',
     api: 'chatCompletions',
     imageHandler: 'novita',
@@ -157,6 +175,7 @@ const providers = {
     outputCapStrategy: 'error'
   },
   openai: {
+    billing: { kind: 'metered' },
     sdk: 'openai',
     apiKeyEnv: 'OPENAI_API_SK',
     createConfiguration: apiKey => ({ apiKey }),
@@ -169,6 +188,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   openrouter: {
+    billing: { kind: 'metered' },
     sdk: 'openrouter',
     apiKeyEnv: 'OPENROUTER_API_SK',
     baseURL: 'https://openrouter.ai/api/v1',
@@ -184,6 +204,7 @@ const providers = {
     }
   },
   qwen: {
+    billing: { kind: 'metered' },
     sdk: 'openai',
     api: 'chatCompletions',
     apiKeyEnv: 'QWEN_API_SK',
@@ -198,6 +219,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   xai: {
+    billing: { kind: 'metered' },
     sdk: 'openai',
     apiKeyEnv: 'XAI_API_SK',
     baseURL: 'https://api.x.ai/v1',
@@ -211,6 +233,7 @@ const providers = {
     outputCapStrategy: 'accept'
   },
   xiaomi: {
+    billing: { kind: 'metered' },
     sdk: 'openai',
     api: 'chatCompletions',
     apiKeyEnv: 'XIAOMI_API_SK',
@@ -223,4 +246,16 @@ const providers = {
 
 Object.freeze(providers)
 
+/**
+ * How calls to `modelId`'s provider are paid for.
+ * @param {string} modelId
+ * @returns {{kind: 'metered' | 'plan' | 'capacity', label?: string, usage?: string}}
+ */
+export function billingOf (modelId) {
+  const provider = providerOf(modelId)
+  const def = providers[provider]
+  if (!def) throw new Error(`Unknown provider '${provider}' in model id '${modelId}'`)
+  return { ...def.billing }
+}
+
 export default providers