mohdel 0.120.0 → 0.121.0

package/README.md

@@ -102,7 +102,7 @@ mo ask anthropic/claude-sonnet-4-6 --stream "write a haiku about recursion"
 mo ask anthropic/claude-opus-4-6 --effort high "prove P != NP"
 
 # On a faster service lane, when the model sells one
-mo ask anthropic/claude-opus-4-6@fast "triage this alert"
+mo ask openai/gpt-5.6-luna@fast "triage this alert"
 
 # Speech → text from an audio file
 mo transcribe groq/whisper-large-v3-turbo meeting.mp3
@@ -318,7 +318,7 @@ What each provider supports through mohdel's unified interface:
 
 Adapter capability ≠ model capability — whether a given model accepts images, tools, or thinking effort depends on the model spec in `curated.json`. The adapter passes through what the envelope supplies; the provider rejects unsupported combos.
 
-**Service speeds** are the exception to that pass-through rule. Where a provider sells the same weights at several speeds (Anthropic's fast mode, and equivalents elsewhere), the lanes a model sells are declared in its `curated.json` entry under `speeds`, with their own prices and rate limits. Select one with `speed` or the `@lane` id suffix. There is no default lane and no fallback: an undeclared lane fails the call before it is sent, because a model that silently ignores an unsupported lane would otherwise be billed at the lane's rates for standard service. See [docs/CATALOG.md](docs/CATALOG.md#service-speeds).
+**Service speeds** are the exception to that pass-through rule. Where a provider sells the same weights at several speeds (OpenAI's `service_tier`, and equivalents elsewhere), the lanes a model sells are declared in its `curated.json` entry under `speeds`, with their own prices and rate limits. Select one with `speed` or the `@lane` id suffix. There is no default lane and no fallback: an undeclared lane fails the call before it is sent, because a model that silently ignores an unsupported lane would otherwise be billed at the lane's rates for standard service. See [docs/CATALOG.md](docs/CATALOG.md#service-speeds).
 
 ## Local Development
 

package/config/curated.schema.json

@@ -2,7 +2,7 @@
   "$schema": "https://json-schema.org/draft/2020-12/schema",
   "$id": "https://github.com/clbrge/mohdel/blob/main/config/curated.schema.json",
   "title": "Mohdel curated.json",
-  "description": "Schema for ~/.config/mohdel/curated.json — see docs/CATALOG.md for the full reference.",
+  "description": "Schema for ~/.config/mohdel/curated.json \u2014 see docs/CATALOG.md for the full reference.",
   "type": "object",
   "propertyNames": {
     "pattern": "^([a-z0-9_-]+/.+|_[a-zA-Z0-9_-]+)$",
@@ -10,16 +10,25 @@
   },
   "additionalProperties": {
     "oneOf": [
-      { "$ref": "#/$defs/deprecatedStub" },
-      { "$ref": "#/$defs/modelEntry" },
-      { "type": "string", "description": "Inline comment value for underscore-prefixed keys." }
+      {
+        "$ref": "#/$defs/deprecatedStub"
+      },
+      {
+        "$ref": "#/$defs/modelEntry"
+      },
+      {
+        "type": "string",
+        "description": "Inline comment value for underscore-prefixed keys."
+      }
     ]
   },
   "$defs": {
     "deprecatedStub": {
       "type": "object",
       "description": "Redirect from a retired model id to its replacement.",
-      "required": ["deprecated"],
+      "required": [
+        "deprecated"
+      ],
       "additionalProperties": true,
       "properties": {
         "deprecated": {
@@ -31,7 +40,11 @@
     "modelEntry": {
       "type": "object",
       "description": "A real model entry. 'model', 'creator', and 'inputFormat' are required.",
-      "required": ["model", "creator", "inputFormat"],
+      "required": [
+        "model",
+        "creator",
+        "inputFormat"
+      ],
       "additionalProperties": true,
       "properties": {
         "model": {
@@ -45,7 +58,15 @@
         "inputFormat": {
           "type": "array",
           "description": "Accepted input modalities. Defaults to ['text'].",
-          "items": { "type": "string", "enum": ["text", "image", "video", "audio"] },
+          "items": {
+            "type": "string",
+            "enum": [
+              "text",
+              "image",
+              "video",
+              "audio"
+            ]
+          },
           "minItems": 1,
           "uniqueItems": true
         },
@@ -53,142 +74,454 @@
           "type": "string",
           "description": "Routing provider. Defaults to the provider segment of the catalog key."
         },
-        "sdk": { "type": "string", "description": "SDK adapter to use (some providers require this)." },
+        "sdk": {
+          "type": "string",
+          "description": "SDK adapter to use (some providers require this)."
+        },
         "type": {
           "type": "string",
-          "enum": ["model", "image", "transcription"],
+          "enum": [
+            "model",
+            "image",
+            "transcription"
+          ],
           "default": "model",
           "description": "'model' for chat/completion, 'image' for image generation, 'transcription' for speech-to-text."
         },
-        "label": { "type": "string", "description": "Human-readable name shown in UIs." },
-        "description": { "type": "string" },
-        "version": { "type": "string" },
-        "createdAt": { "type": "string", "format": "date-time" },
-        "created": { "type": "number", "description": "Unix timestamp (seconds)." },
-
+        "label": {
+          "type": "string",
+          "description": "Human-readable name shown in UIs."
+        },
+        "description": {
+          "type": "string"
+        },
+        "version": {
+          "type": "string"
+        },
+        "createdAt": {
+          "type": "string",
+          "format": "date-time"
+        },
+        "created": {
+          "type": "number",
+          "description": "Unix timestamp (seconds)."
+        },
         "inputPrice": {
           "oneOf": [
-            { "type": "number", "minimum": 0 },
-            { "type": "object", "required": ["default"], "additionalProperties": { "type": "number" }, "properties": { "default": { "type": "number" } } }
+            {
+              "type": "number",
+              "minimum": 0
+            },
+            {
+              "type": "object",
+              "required": [
+                "default"
+              ],
+              "additionalProperties": {
+                "type": "number"
+              },
+              "properties": {
+                "default": {
+                  "type": "number"
+                }
+              }
+            }
           ],
-          "description": "USD per 1M input tokens. Object form is for tiered pricing — must include a 'default' key."
+          "description": "USD per 1M input tokens. Object form is for tiered pricing \u2014 must include a 'default' key."
         },
         "outputPrice": {
           "oneOf": [
-            { "type": "number", "minimum": 0 },
-            { "type": "object", "required": ["default"], "additionalProperties": { "type": "number" }, "properties": { "default": { "type": "number" } } }
+            {
+              "type": "number",
+              "minimum": 0
+            },
+            {
+              "type": "object",
+              "required": [
+                "default"
+              ],
+              "additionalProperties": {
+                "type": "number"
+              },
+              "properties": {
+                "default": {
+                  "type": "number"
+                }
+              }
+            }
           ],
           "description": "USD per 1M output tokens."
         },
         "thinkingPrice": {
           "oneOf": [
-            { "type": "number", "minimum": 0 },
-            { "type": "object", "required": ["default"], "additionalProperties": { "type": "number" }, "properties": { "default": { "type": "number" } } }
+            {
+              "type": "number",
+              "minimum": 0
+            },
+            {
+              "type": "object",
+              "required": [
+                "default"
+              ],
+              "additionalProperties": {
+                "type": "number"
+              },
+              "properties": {
+                "default": {
+                  "type": "number"
+                }
+              }
+            }
           ],
           "description": "USD per 1M thinking/reasoning tokens (when the provider bills these separately)."
         },
-        "cacheWritePrice": { "type": "number", "minimum": 0, "description": "USD per 1M tokens written to provider-side prompt cache." },
-        "cacheWrite1hPrice": { "type": "number", "minimum": 0, "description": "USD per 1M tokens written with a 1h TTL, when the provider prices that above the default write rate. Falls back to cacheWritePrice." },
-        "cacheReadPrice": { "type": "number", "minimum": 0, "description": "USD per 1M tokens served from provider-side prompt cache." },
-
-        "contextTokenLimit": { "type": "integer", "minimum": 1, "description": "Maximum total tokens (input + output)." },
-        "outputTokenLimit": { "type": "integer", "minimum": 1, "description": "Maximum output tokens per call." },
-        "thinkingTokenLimit": { "type": "integer", "minimum": 1, "description": "Maximum thinking tokens per call (when separate from output)." },
-        "tokenizerHeadroom": { "type": "number", "exclusiveMinimum": 0, "description": "Multiplier applied to local token estimates to account for tokenizer drift." },
-
+        "cacheWritePrice": {
+          "oneOf": [
+            {
+              "type": "number",
+              "minimum": 0
+            },
+            {
+              "type": "object",
+              "required": [
+                "default"
+              ],
+              "additionalProperties": {
+                "type": "number"
+              },
+              "properties": {
+                "default": {
+                  "type": "number"
+                }
+              }
+            }
+          ],
+          "description": "USD per 1M tokens written to provider-side prompt cache. Object form is tiered on input tokens."
+        },
+        "cacheWrite1hPrice": {
+          "oneOf": [
+            {
+              "type": "number",
+              "minimum": 0
+            },
+            {
+              "type": "object",
+              "required": [
+                "default"
+              ],
+              "additionalProperties": {
+                "type": "number"
+              },
+              "properties": {
+                "default": {
+                  "type": "number"
+                }
+              }
+            }
+          ],
+          "description": "USD per 1M tokens written with a 1h TTL, when the provider prices that above the default write rate. Falls back to cacheWritePrice."
+        },
+        "cacheReadPrice": {
+          "oneOf": [
+            {
+              "type": "number",
+              "minimum": 0
+            },
+            {
+              "type": "object",
+              "required": [
+                "default"
+              ],
+              "additionalProperties": {
+                "type": "number"
+              },
+              "properties": {
+                "default": {
+                  "type": "number"
+                }
+              }
+            }
+          ],
+          "description": "USD per 1M tokens served from provider-side prompt cache. Object form is tiered on input tokens."
+        },
+        "contextTokenLimit": {
+          "type": "integer",
+          "minimum": 1,
+          "description": "Maximum total tokens (input + output)."
+        },
+        "outputTokenLimit": {
+          "type": "integer",
+          "minimum": 1,
+          "description": "Maximum output tokens per call."
+        },
+        "thinkingTokenLimit": {
+          "type": "integer",
+          "minimum": 1,
+          "description": "Maximum thinking tokens per call (when separate from output)."
+        },
+        "tokenizerHeadroom": {
+          "type": "number",
+          "exclusiveMinimum": 0,
+          "description": "Multiplier applied to local token estimates to account for tokenizer drift."
+        },
         "thinkingEffortLevels": {
           "oneOf": [
-            { "type": "null" },
+            {
+              "type": "null"
+            },
             {
               "type": "object",
-              "description": "Map effort name → provider-native budget. Standard names: low/medium/high/xhigh/max/none.",
-              "additionalProperties": { "type": "number", "minimum": 0 }
+              "description": "Map effort name \u2192 provider-native budget. Standard names: low/medium/high/xhigh/max/none.",
+              "additionalProperties": {
+                "type": "number",
+                "minimum": 0
+              }
             }
           ]
         },
-        "defaultThinkingEffort": { "type": "string", "description": "Effort level used when the envelope omits 'outputEffort'." },
-
+        "defaultThinkingEffort": {
+          "type": "string",
+          "description": "Effort level used when the envelope omits 'outputEffort'."
+        },
         "speeds": {
           "type": "object",
-          "description": "Service speed lanes this model sells, keyed by lane name. A lane is a provider request parameter buying a different speed/price point for the same weights. There is no default lane: omitting 'speed' on the envelope sends no parameter. Only lanes declared here may be requested.",
+          "description": "Service speed lanes this model sells, keyed by lane name. A lane is a provider request parameter buying a different speed/price point for the same weights. There is no default lane: omitting 'speed' on the envelope sends no parameter. Only lanes declared here may be requested, and only names the provider's adapter accepts.",
           "additionalProperties": {
             "type": "object",
-            "required": ["wire"],
             "additionalProperties": false,
-            "description": "Lane overlay. Fields named here override the base entry; unnamed fields fall through. Only prices and rate limits are overridable — anything that would change what the model is belongs in its own entry.",
+            "description": "Lane overlay: the price and rate-limit fields that differ from the base entry. Unnamed fields fall through. Lane names come from the provider's own vocabulary (OpenAI: fast, priority, flex, scale) \u2014 the adapter decides how a name reaches the wire.",
             "properties": {
-              "wire": { "type": "string", "description": "Provider-native value for the lane parameter (e.g. 'fast')." },
               "inputPrice": {
                 "oneOf": [
-                  { "type": "number", "minimum": 0 },
-                  { "type": "object", "required": ["default"], "additionalProperties": { "type": "number" }, "properties": { "default": { "type": "number" } } }
+                  {
+                    "type": "number",
+                    "minimum": 0
+                  },
+                  {
+                    "type": "object",
+                    "required": [
+                      "default"
+                    ],
+                    "additionalProperties": {
+                      "type": "number"
+                    },
+                    "properties": {
+                      "default": {
+                        "type": "number"
+                      }
+                    }
+                  }
                 ]
               },
               "outputPrice": {
                 "oneOf": [
-                  { "type": "number", "minimum": 0 },
-                  { "type": "object", "required": ["default"], "additionalProperties": { "type": "number" }, "properties": { "default": { "type": "number" } } }
+                  {
+                    "type": "number",
+                    "minimum": 0
+                  },
+                  {
+                    "type": "object",
+                    "required": [
+                      "default"
+                    ],
+                    "additionalProperties": {
+                      "type": "number"
+                    },
+                    "properties": {
+                      "default": {
+                        "type": "number"
+                      }
+                    }
+                  }
                 ]
               },
               "thinkingPrice": {
                 "oneOf": [
-                  { "type": "number", "minimum": 0 },
-                  { "type": "object", "required": ["default"], "additionalProperties": { "type": "number" }, "properties": { "default": { "type": "number" } } }
+                  {
+                    "type": "number",
+                    "minimum": 0
+                  },
+                  {
+                    "type": "object",
+                    "required": [
+                      "default"
+                    ],
+                    "additionalProperties": {
+                      "type": "number"
+                    },
+                    "properties": {
+                      "default": {
+                        "type": "number"
+                      }
+                    }
+                  }
+                ]
+              },
+              "cacheWritePrice": {
+                "oneOf": [
+                  {
+                    "type": "number",
+                    "minimum": 0
+                  },
+                  {
+                    "type": "object",
+                    "required": [
+                      "default"
+                    ],
+                    "additionalProperties": {
+                      "type": "number"
+                    },
+                    "properties": {
+                      "default": {
+                        "type": "number"
+                      }
+                    }
+                  }
                 ]
               },
-              "cacheWritePrice": { "type": "number", "minimum": 0 },
-              "cacheWrite1hPrice": { "type": "number", "minimum": 0 },
-              "cacheReadPrice": { "type": "number", "minimum": 0 },
-              "rpmLimit": { "type": "integer", "minimum": 1 },
-              "tpmLimit": { "type": "integer", "minimum": 1 }
+              "cacheWrite1hPrice": {
+                "oneOf": [
+                  {
+                    "type": "number",
+                    "minimum": 0
+                  },
+                  {
+                    "type": "object",
+                    "required": [
+                      "default"
+                    ],
+                    "additionalProperties": {
+                      "type": "number"
+                    },
+                    "properties": {
+                      "default": {
+                        "type": "number"
+                      }
+                    }
+                  }
+                ]
+              },
+              "cacheReadPrice": {
+                "oneOf": [
+                  {
+                    "type": "number",
+                    "minimum": 0
+                  },
+                  {
+                    "type": "object",
+                    "required": [
+                      "default"
+                    ],
+                    "additionalProperties": {
+                      "type": "number"
+                    },
+                    "properties": {
+                      "default": {
+                        "type": "number"
+                      }
+                    }
+                  }
+                ]
+              },
+              "rpmLimit": {
+                "type": "integer",
+                "minimum": 1
+              },
+              "tpmLimit": {
+                "type": "integer",
+                "minimum": 1
+              }
             }
           }
         },
-
         "tags": {
           "type": "array",
-          "items": { "type": "string", "pattern": "^[a-zA-Z][a-zA-Z0-9._-]{0,31}$" },
+          "items": {
+            "type": "string",
+            "pattern": "^[a-zA-Z][a-zA-Z0-9._-]{0,31}$"
+          },
           "uniqueItems": true,
           "description": "Free-form classification tags (used by 'mo bench --tag', 'mo rank', and caller-side selection)."
         },
         "aliases": {
           "type": "array",
-          "items": { "type": "string" },
+          "items": {
+            "type": "string"
+          },
           "description": "Alternative ids that should resolve to this entry."
         },
         "replaces": {
           "type": "array",
-          "items": { "type": "string" },
+          "items": {
+            "type": "string"
+          },
           "description": "Older model ids this one supersedes."
         },
         "leaderboard": {
           "type": "array",
-          "items": { "type": "number" },
+          "items": {
+            "type": "number"
+          },
           "minItems": 3,
           "maxItems": 3,
-          "description": "[intelligence, speed, latency] triple — drives 'mo rank'."
-        },
-        "leaderboardNote": { "type": "string" },
-
-        "supportsTools": { "type": "boolean", "description": "Set false to mark a model as tool-less." },
-
-        "imagePrice": { "type": "number", "minimum": 0, "description": "USD per generated image (image-type entries)." },
-        "imageEndpoint": { "type": "string", "description": "Provider-side image endpoint name." },
-        "imageDefaultSize": { "type": "string", "description": "Default size when envelope omits one (e.g. '1024x1024')." },
-        "transcriptionPrice": { "type": "number", "minimum": 0, "description": "USD per audio minute (transcription-type entries). Token-billed transcription models (OpenAI gpt-4o-*-transcribe) use inputPrice/outputPrice instead." },
-
-        "rpmLimit": { "type": "integer", "minimum": 1, "description": "Requests per minute. Overrides provider default." },
-        "tpmLimit": { "type": "integer", "minimum": 1, "description": "Tokens per minute. Overrides provider default." },
+          "description": "[intelligence, speed, latency] triple \u2014 drives 'mo rank'."
+        },
+        "leaderboardNote": {
+          "type": "string"
+        },
+        "supportsTools": {
+          "type": "boolean",
+          "description": "Set false to mark a model as tool-less."
+        },
+        "imagePrice": {
+          "type": "number",
+          "minimum": 0,
+          "description": "USD per generated image (image-type entries)."
+        },
+        "imageEndpoint": {
+          "type": "string",
+          "description": "Provider-side image endpoint name."
+        },
+        "imageDefaultSize": {
+          "type": "string",
+          "description": "Default size when envelope omits one (e.g. '1024x1024')."
+        },
+        "transcriptionPrice": {
+          "type": "number",
+          "minimum": 0,
+          "description": "USD per audio minute (transcription-type entries). Token-billed transcription models (OpenAI gpt-4o-*-transcribe) use inputPrice/outputPrice instead."
+        },
+        "rpmLimit": {
+          "type": "integer",
+          "minimum": 1,
+          "description": "Requests per minute. Overrides provider default."
+        },
+        "tpmLimit": {
+          "type": "integer",
+          "minimum": 1,
+          "description": "Tokens per minute. Overrides provider default."
+        },
         "rateLimitScope": {
           "type": "string",
-          "enum": ["model", "provider"],
+          "enum": [
+            "model",
+            "provider"
+          ],
           "description": "'model' = private budget. 'provider' = shared with provider-level pool."
         },
-
-        "deprecated": { "type": "string", "description": "If present, this entry is treated as a stub (use 'deprecatedStub' shape)." },
-        "suspended": { "type": "string", "description": "Reason this model is temporarily disabled." },
-
-        "displayName": { "type": "string", "deprecated": true, "description": "Deprecated — use 'label'." }
+        "deprecated": {
+          "type": "string",
+          "description": "If present, this entry is treated as a stub (use 'deprecatedStub' shape)."
+        },
+        "suspended": {
+          "type": "string",
+          "description": "Reason this model is temporarily disabled."
+        },
+        "displayName": {
+          "type": "string",
+          "deprecated": true,
+          "description": "Deprecated \u2014 use 'label'."
+        }
       }
     }
   }

package/js/session/adapters/_chat_completions.js

@@ -19,7 +19,6 @@
 import { getSpec } from './_catalog.js'
 import { classifyProviderError } from './_errors.js'
 import { costFor } from './_pricing.js'
-import { applySpeed } from './_speed.js'
 import { catalogKey, bareOf } from '#core/model-id.js'
 import {
   STATUS_COMPLETED,
@@ -371,8 +370,6 @@ function buildRequest (envelope, spec, config) {
     args[config.identifierField || 'user'] = envelope.identifier
   }
 
-  applySpeed(args, envelope, spec)
-
   return args
 }
 

package/js/session/adapters/_speed.js

@@ -1,30 +1,21 @@
 /**
- * Service-speed lanes.
+ * Service-speed lanes — catalog side.
  *
- * A lane is a provider request parameter that buys a different
- * speed/price point for the same weights (Anthropic `speed: "fast"`).
- * Lanes are declared per catalog entry under `speeds`, keyed by lane
- * name, each carrying the wire value plus any price and rate-limit
- * fields that differ from the base entry.
+ * A lane is a named service speed a model sells, declared per catalog
+ * entry under `speeds` with the price and rate-limit fields that
+ * differ from the base entry. This module owns only what is common to
+ * every provider: what the entry declares, and what that means for
+ * pricing and throttling.
  *
- * Two declarations must agree for a lane to be usable: the entry
- * declares it (`spec.speeds`) and the provider's adapter can emit it
- * (`SPEED_PARAMS`). `run.js` checks both before dispatch.
+ * How a lane reaches the wire, and how the served lane is read back,
+ * is provider protocol and lives in the adapter. An adapter that
+ * handles lanes advertises the names it accepts on `speedLanes`;
+ * absence of that property is the declaration that it handles none,
+ * and is what `run.js` checks before dispatch.
  *
  * @module session/adapters/_speed
  */
 
-import { providerOf } from '#core/model-id.js'
-
-/**
- * Providers whose adapter emits a lane parameter, mapped to the
- * native parameter name. Absence from this table is the declaration
- * that a provider has no lanes.
- */
-export const SPEED_PARAMS = Object.freeze({
-  anthropic: 'speed'
-})
-
 /** Spec fields a lane overlay may restate. */
 export const SPEED_OVERRIDABLE = Object.freeze([
   'inputPrice',
@@ -37,25 +28,9 @@ export const SPEED_OVERRIDABLE = Object.freeze([
   'tpmLimit'
 ])
 
-/**
- * @param {string} provider
- * @returns {boolean}
- */
-export function providerSupportsSpeed (provider) {
-  return Object.hasOwn(SPEED_PARAMS, provider)
-}
-
-/**
- * @param {string} provider
- * @returns {string | undefined}
- */
-export function speedParamFor (provider) {
-  return SPEED_PARAMS[provider]
-}
-
 /**
  * Whether `spec` declares `speed`. Uses `hasOwn` rather than
- * truthiness so an overlay that only carries `wire` still counts.
+ * truthiness so a lane sold at base prices still counts.
  *
  * @param {any} spec
  * @param {string} speed
@@ -73,6 +48,22 @@ export function speedNames (spec) {
   return spec?.speeds ? Object.keys(spec.speeds) : []
 }
 
+/**
+ * Whether the lane declares a quota of its own, which is what earns it
+ * a private rate-limit bucket. Lanes that don't (OpenAI's service_tier
+ * shares the model's TPM/RPM pool) count against the base bucket —
+ * giving them their own would silently double the allowance.
+ *
+ * @param {any} spec
+ * @param {string} [speed]
+ * @returns {boolean}
+ */
+export function speedHasOwnQuota (spec, speed) {
+  if (!speed || !hasSpeed(spec, speed)) return false
+  const overlay = spec.speeds[speed]
+  return overlay.rpmLimit != null || overlay.tpmLimit != null
+}
+
 /**
  * Spec with `speed`'s overlay applied. Unnamed fields fall through to
  * the base entry.
@@ -94,26 +85,3 @@ export function mergeSpeed (spec, speed) {
   }
   return merged
 }
-
-/**
- * Set the provider-native lane parameter on an outbound request.
- * No-op when the envelope carries no lane.
- *
- * @param {Record<string, any>} request
- * @param {import('#core/envelope.js').CallEnvelope} envelope
- * @param {any} spec
- * @throws when the envelope carries a lane the provider cannot emit
- *   or the spec does not declare
- */
-export function applySpeed (request, envelope, spec) {
-  if (!envelope.speed) return
-  const provider = providerOf(envelope.model)
-  const param = speedParamFor(provider)
-  if (!param) {
-    throw new Error(`provider '${provider}' does not implement speed lanes`)
-  }
-  if (!hasSpeed(spec, envelope.speed)) {
-    throw new Error(`model does not declare speed lane '${envelope.speed}'`)
-  }
-  request[param] = spec.speeds[envelope.speed].wire
-}

package/js/session/adapters/anthropic.js

@@ -29,7 +29,6 @@ import { classifyProviderError } from './_errors.js'
 import { loadImages } from './_images.js'
 import { isTrustedMedia } from './_media.js'
 import { costFor } from './_pricing.js'
-import { applySpeed } from './_speed.js'
 import { catalogKey, bareOf } from '#core/model-id.js'
 import {
   toAnthropicTools,
@@ -353,7 +352,6 @@ function buildRequest (envelope, conversation, system, conversationCacheTtl = nu
     }
   }
 
-  applySpeed(request, envelope, spec)
   applyCacheBreakpoints(request, conversationCacheTtl)
 
   return request

package/js/session/adapters/gemini.js

@@ -31,7 +31,6 @@ import { loadImages } from './_images.js'
 import { isTrustedMedia } from './_media.js'
 import { loadVideos } from './_videos.js'
 import { costFor } from './_pricing.js'
-import { applySpeed } from './_speed.js'
 import { catalogKey, bareOf } from '#core/model-id.js'
 import {
   toGeminiTools,
@@ -272,7 +271,6 @@ function buildRequest (envelope, contents, systemInstruction) {
     contents
   }
   if (Object.keys(config).length > 0) request.config = config
-  applySpeed(request, envelope, spec)
   return request
 }
 

package/js/session/adapters/openai.js

@@ -30,7 +30,6 @@ import { classifyProviderError } from './_errors.js'
 import { loadImages } from './_images.js'
 import { isTrustedMedia } from './_media.js'
 import { costFor } from './_pricing.js'
-import { applySpeed } from './_speed.js'
 import { catalogKey, providerOf, bareOf } from '#core/model-id.js'
 import {
   toOpenAITools,
@@ -84,6 +83,8 @@ export async function * openai (envelope, deps = {}) {
   let status = STATUS_COMPLETED
   /** @type {string | undefined} */
   let warning
+  /** @type {string | null | undefined} */
+  let servedTier
 
   // Tool-call accumulation: itemId → {call_id, name, arguments}
   /** @type {Map<string, {call_id: string, name: string, arguments: string}>} */
@@ -131,6 +132,7 @@ export async function * openai (envelope, deps = {}) {
           break
 
         case 'response.completed':
+          servedTier = event.response?.service_tier ?? servedTier
           if (event.response?.usage) {
             inputTokens = event.response.usage.input_tokens ?? 0
             outputTokens = event.response.usage.output_tokens ?? 0
@@ -144,6 +146,7 @@ export async function * openai (envelope, deps = {}) {
           break
 
         case 'response.incomplete':
+          servedTier = event.response?.service_tier ?? servedTier
           status = STATUS_INCOMPLETE
           if (event.response?.incomplete_details?.reason === 'max_output_tokens') {
             warning = WARNING_INSUFFICIENT_OUTPUT_BUDGET
@@ -190,6 +193,8 @@ export async function * openai (envelope, deps = {}) {
   // simpler with the additive shape.
   const regularInputTokens = Math.max(0, inputTokens - cachedInputTokens - cacheWriteTokens)
 
+  const billed = billedEnvelope(envelope, servedTier, log)
+
   /** @type {import('#core/events.js').DoneEvent} */
   const done = {
     type: 'done',
@@ -201,8 +206,9 @@ export async function * openai (envelope, deps = {}) {
       thinkingTokens,
       ...(cacheWriteTokens > 0 && { cacheWriteInputTokens: cacheWriteTokens }),
       ...(cachedInputTokens > 0 && { cacheReadInputTokens: cachedInputTokens }),
+      ...(envelope.speed && { servedSpeed: billed.served }),
       cost: costFor(
-        envelope,
+        billed.envelope,
         {
           inputTokens: regularInputTokens,
           outputTokens: messageOutputTokens,
@@ -221,6 +227,66 @@ export async function * openai (envelope, deps = {}) {
   yield done
 }
 
+/**
+ * Lane names the OpenAI adapter can put on `service_tier`. `run.js`
+ * reads this before dispatch; a lane outside it never reaches here.
+ */
+openai.speedLanes = new Set(['fast', 'priority', 'flex', 'scale'])
+
+/**
+ * Lane the response says was served, given the one requested.
+ *
+ * OpenAI answers `service_tier: 'priority'` to a granted request for
+ * either premium lane, so the echo alone cannot name which was asked
+ * for — the request is the other half of the answer. Any other value
+ * is the tier that actually ran, `null` when nothing was reported.
+ *
+ * @param {string} requested
+ * @param {string | null | undefined} servedTier
+ * @returns {string | null | undefined}
+ */
+function servedLane (requested, servedTier) {
+  if (servedTier == null) return undefined
+  if (servedTier === 'priority') {
+    return (requested === 'fast' || requested === 'priority') ? requested : 'priority'
+  }
+  return openai.speedLanes.has(servedTier) ? servedTier : null
+}
+
+/**
+ * Envelope to price the call against. OpenAI serves Standard instead
+ * when a premium lane is unavailable — documented behaviour above the
+ * ramp rate limit — so billing the requested lane would charge premium
+ * rates for standard service.
+ *
+ * When a lane was requested and nothing was reported back, the request
+ * is the only evidence available: bill it and say so.
+ *
+ * @param {import('#core/envelope.js').CallEnvelope} envelope
+ * @param {string | null | undefined} servedTier
+ * @param {any} [log]
+ * @returns {{envelope: import('#core/envelope.js').CallEnvelope, served: string | null}}
+ */
+function billedEnvelope (envelope, servedTier, log) {
+  if (!envelope.speed) return { envelope, served: null }
+
+  const served = servedLane(envelope.speed, servedTier)
+  if (served === undefined) {
+    log?.warn(
+      { model: envelope.model, speed: envelope.speed },
+      '[mohdel:openai] no service_tier reported; billing the requested lane'
+    )
+    return { envelope, served: envelope.speed }
+  }
+  if (served === envelope.speed) return { envelope, served }
+
+  log?.warn(
+    { model: envelope.model, requested: envelope.speed, servedTier, served },
+    '[mohdel:openai] provider served a different speed lane; billing what was served'
+  )
+  return { envelope: { ...envelope, speed: served ?? undefined }, served }
+}
+
 /**
  * @param {import('#core/envelope.js').CallEnvelope} envelope
  * @param {Array<any>} input
@@ -288,7 +354,7 @@ function buildRequest (envelope, input, instructions) {
     }
   }
 
-  applySpeed(request, envelope, spec)
+  if (envelope.speed) request.service_tier = envelope.speed
 
   return request
 }

package/js/session/run.js

@@ -26,7 +26,7 @@ import { getAdapter } from './adapters/index.js'
 import { isImageProvider } from './adapters/image/index.js'
 import { getSpec } from './adapters/_catalog.js'
 import { getProviderLimits } from './adapters/_providers.js'
-import { hasSpeed, mergeSpeed, providerSupportsSpeed, speedNames } from './adapters/_speed.js'
+import { hasSpeed, mergeSpeed, speedHasOwnQuota, speedNames } from './adapters/_speed.js'
 import { providerOf, catalogKey, effortOf, speedOf } from '#core/model-id.js'
 import * as defaultCooldown from './_cooldown.js'
 import * as defaultLimiter from './_rate_limiter.js'
@@ -128,7 +128,7 @@ export async function * run (envelope, {
   }
 
   if (envelope.speed) {
-    const speedErr = speedError(key, envelope.speed, spec, provider)
+    const speedErr = speedError(key, envelope.speed, spec, provider, adapter)
     if (speedErr) {
       log.warn({ provider, speed: envelope.speed }, '[mohdel:answer] unusable speed lane')
       endSpanError(span, new Error(speedErr.error.message))
@@ -150,12 +150,10 @@ export async function * run (envelope, {
   const providerCfg = resolveProviderLimits(provider) || {}
   const rpmLimit = effective?.rpmLimit ?? providerCfg.rpmLimit
   const tpmLimit = effective?.tpmLimit ?? providerCfg.tpmLimit
-  // An active lane is its own capacity pool, so it gets its own bucket
-  // whatever `rateLimitScope` says — sharing one would let standard
-  // traffic throttle the lane being paid for.
-  const bucketKey = envelope.speed
+  const baseBucket = (spec?.rateLimitScope === 'model') ? key : provider
+  const bucketKey = speedHasOwnQuota(spec, envelope.speed)
     ? `${key}@${envelope.speed}`
-    : (spec?.rateLimitScope === 'model' ? key : provider)
+    : baseBucket
 
   // `0` is a killswitch ("deny all"), not "unset"; `undefined`/`null`
   // means no limit configured for that dimension. Gate on nullability
@@ -363,25 +361,31 @@ function normalizeModelId (envelope, resolveSpec) {
  * @param {string} speed
  * @param {any} spec
  * @param {string} provider
+ * @param {any} adapter
  * @returns {import('#core/events.js').ErrorEvent | undefined}
  */
-function speedError (key, speed, spec, provider) {
+function speedError (key, speed, spec, provider, adapter) {
   if (!hasSpeed(spec, speed)) {
     const available = speedNames(spec)
     const detail = available.length
       ? `Available: ${available.join(', ')}`
       : 'It declares no speed lanes.'
-    const hint = speed.includes(':')
-      ? " Suffix order is ':effort' then '@speed'."
-      : ''
+    const colon = speed.indexOf(':')
+    const hint = colon < 0
+      ? ''
+      : ` Suffix order is ':effort' then '@speed' — did you mean '${key}:${speed.slice(colon + 1)}@${speed.slice(0, colon)}'?`
     return errorEvent(
       `Model '${key}' does not support speed lane '${speed}'. ${detail}${hint}`,
       'SESSION_INVALID_SPEED'
     )
   }
-  if (!providerSupportsSpeed(provider)) {
+  const lanes = adapter?.speedLanes
+  if (!lanes?.has(speed)) {
+    const detail = lanes
+      ? `It accepts: ${[...lanes].join(', ')}.`
+      : 'It implements no speed lanes.'
     return errorEvent(
-      `Provider '${provider}' does not implement speed lanes, but '${key}' declares '${speed}'.`,
+      `Provider '${provider}' cannot serve speed lane '${speed}', but '${key}' declares it. ${detail}`,
       'SESSION_SPEED_NOT_IMPLEMENTED'
     )
   }
@@ -456,6 +460,8 @@ function finalizeSpanOk (span, result, sawDelta = false, maxInterFrameMs = 0) {
   }
   if (result?.cacheWriteInputTokens) attrs['mohdel.cache_write_input_tokens'] = result.cacheWriteInputTokens
   if (result?.cacheReadInputTokens) attrs['mohdel.cache_read_input_tokens'] = result.cacheReadInputTokens
+  if (result?.speed) attrs['mohdel.speed'] = result.speed
+  if (result?.servedSpeed !== undefined) attrs['mohdel.served_speed'] = result.servedSpeed ?? 'standard'
   if (result?.cost != null) attrs['mohdel.cost'] = result.cost
   if (result?.warning) attrs['mohdel.warning'] = result.warning
   if (result?.timestamps?.start && result?.timestamps?.first) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mohdel",
-  "version": "0.120.0",
+  "version": "0.121.0",
   "license": "MIT",
   "author": {
     "name": "Christophe Le Bars",
@@ -108,12 +108,12 @@
     "@opentelemetry/exporter-trace-otlp-grpc": "^0.221.0",
     "@opentelemetry/sdk-node": "^0.221.0",
     "chalk": "^6.0.0",
-    "mohdel-thin-gate-linux-x64-gnu": "0.120.0"
+    "mohdel-thin-gate-linux-x64-gnu": "0.121.0"
   },
   "dependencies": {
-    "@anthropic-ai/sdk": "^0.115.0",
+    "@anthropic-ai/sdk": "^0.117.1",
     "@cerebras/cerebras_cloud_sdk": "^1.91.0",
-    "@google/genai": "^2.16.0",
+    "@google/genai": "^2.17.1",
     "@opentelemetry/api": "^1.9.1",
     "env-paths": "^4.0.0",
     "groq-sdk": "^1.5.0",
@@ -126,7 +126,7 @@
   "devDependencies": {
     "gpt-tokenizer": "^3.4.0",
     "lint-staged": "^17.3.0",
-    "release-it": "^21.0.1",
+    "release-it": "^21.0.2",
     "standard": "^17.1.2",
     "vitest": "^4.1.10"
   }

package/src/cli/ask.js

@@ -176,6 +176,12 @@ Examples:
     if (tokens.outputTokens) summary.push(`${tokens.outputTokens} out`)
     if (tokens.thinkingTokens) summary.push(`${tokens.thinkingTokens} think`)
     if (tokens.cost != null) summary.push(`$${tokens.cost.toFixed(4)}`)
+    if (tokens.speed) {
+      const served = tokens.servedSpeed
+      summary.push(served === tokens.speed
+        ? `${tokens.speed} lane`
+        : `${tokens.speed} lane → ${served ?? 'standard'}`)
+    }
     const ts = tokens.timestamps
     if (ts) {
       const toMs = (a, b) => {

package/src/cli/check.js

@@ -1,7 +1,7 @@
 import { label, err, warn, ok } from './colors.js'
 import providers from '../lib/providers.js'
 import { validate, isValidTag } from '../lib/schema.js'
-import { providerSupportsSpeed } from '../../js/session/adapters/_speed.js'
+import { adapters } from '../../js/session/adapters/index.js'
 import { getCuratedModels, loadDefaultEnv, catalogEntries, catalogValues } from '../lib/common.js'
 
 // --- Local validation ---
@@ -51,10 +51,12 @@ const checkLocal = (curated) => {
       warnings.push(`${key}: has thinkingEffortLevels but no defaultThinkingEffort`)
     }
 
-    if (spec.speeds && !providerSupportsSpeed(keyProvider)) {
-      errors.push(`${key}: declares speeds but provider '${keyProvider}' has no adapter support — calls on those lanes would fail at dispatch`)
-    }
+    const lanes = adapters[keyProvider]?.speedLanes
     for (const [lane, overlay] of Object.entries(spec.speeds || {})) {
+      if (!lanes?.has(lane)) {
+        const detail = lanes ? `accepts: ${[...lanes].join(', ')}` : 'implements no speed lanes'
+        errors.push(`${key}: speeds.${lane} — provider '${keyProvider}' ${detail}; calls on that lane would fail at dispatch`)
+      }
       for (const priceField of ['inputPrice', 'outputPrice', 'thinkingPrice']) {
         const val = overlay[priceField]
         if (val != null && typeof val === 'object' && val.default == null) {

package/src/lib/index.js

@@ -75,6 +75,17 @@ const resolvePrice = (price, inputTokens) => {
 // @internal — exported for unit tests only.
 export { resolvePrice as _resolvePriceForTests }
 
+// A lane candidate containing `:` means the suffixes were written the
+// other way round, which is otherwise reported as an unknown lane
+// spelled `fast:none`.
+const reversedSuffixHint = (base, candidate) => {
+  const colon = candidate.indexOf(':')
+  if (colon < 0) return ''
+  const speed = candidate.slice(0, colon)
+  const effort = candidate.slice(colon + 1)
+  return ` Suffix order is ':effort' then '@speed' — did you mean '${base}:${effort}@${speed}'?`
+}
+
 const normalizeModelSpec = (resolvedModelId, modelSpec, providerConfig) => {
   const normalized = { ...modelSpec }
   if (!normalized.provider) {
@@ -357,9 +368,12 @@ const mohdel = async ({ logger, verbosity: verbosityOpt, onSuccess, onFailure, c
           if (aliasSpeed && !Object.hasOwn(modelSpec.speeds || {}, aliasSpeed)) {
             const available = Object.keys(modelSpec.speeds || {})
             const detail = available.length
-              ? `Available: ${available.join(', ')}`
+              ? `Available: ${available.join(', ')}.`
               : 'It declares no speed lanes.'
-            throw new Error(`Model '${resolvedModelId}' does not support speed lane '${aliasSpeed}'. ${detail}`)
+            throw new Error(
+              `Model '${resolvedModelId}' does not support speed lane '${aliasSpeed}'. ${detail}` +
+              reversedSuffixHint(resolvedModelId, aliasSpeed)
+            )
           }
 
           // Validate outputEffort alias against model capabilities

package/src/lib/schema.js

@@ -1,14 +1,11 @@
 import { SPEED_OVERRIDABLE } from '../../js/session/adapters/_speed.js'
 
 const validateSpeeds = (speeds) => {
-  const allowed = new Set(['wire', ...SPEED_OVERRIDABLE])
+  const allowed = new Set(SPEED_OVERRIDABLE)
   for (const [lane, overlay] of Object.entries(speeds)) {
     if (typeof overlay !== 'object' || overlay === null || Array.isArray(overlay)) {
       return `lane '${lane}' must be an object`
     }
-    if (typeof overlay.wire !== 'string' || !overlay.wire) {
-      return `lane '${lane}' must set a string 'wire' value`
-    }
     for (const field of Object.keys(overlay)) {
       if (!allowed.has(field)) {
         return `lane '${lane}' may not override '${field}' (allowed: ${[...allowed].join(', ')})`
@@ -30,6 +27,9 @@ const fieldDefs = {
   inputPrice: { type: 'number', altType: 'object' },
   outputPrice: { type: 'number', altType: 'object' },
   thinkingPrice: { type: 'number', altType: 'object' },
+  cacheReadPrice: { type: 'number', altType: 'object' },
+  cacheWritePrice: { type: 'number', altType: 'object' },
+  cacheWrite1hPrice: { type: 'number', altType: 'object' },
   contextTokenLimit: { type: 'number' },
   outputTokenLimit: { type: 'number' },
   thinkingTokenLimit: { type: 'number' },