mohdel 3.2.0 → 3.4.0

package/README.md

@@ -27,7 +27,7 @@ Mohdel runs the inference layer of production stacks, among them [docAnalyzer](h
 ## Why mohdel
 
 - **Real numbers on every call.** Token counts and per-call USD cost computed from your own pricing catalog (`curated.json`) — not estimates, not provider-specific shapes. Bill tenants, alert on spend, reconcile invoices. Your own catalog means your negotiated rates and your own tags, and it is not a spreadsheet you maintain: `mo model instructions` hands the provider's docs page to your coding agent, which drafts the entries for you to review. See [docs/CATALOG.md](docs/CATALOG.md).
-- **One interface across providers.** Same `answer()` call, same event stream, same `{ status, output, inputTokens, outputTokens, cost }` result. Switching from `anthropic/claude-sonnet-4-6` to `openai/gpt-5.4-mini` is one string change — adapter differences stay inside mohdel.
+- **One interface across providers.** Same `answer()` call, same event stream, same `{ status, output, inputTokens, outputTokens, cost }` result. Switching from `anthropic/claude-sonnet-5-5` to `openai/gpt-5.4-mini` is one string change — adapter differences stay inside mohdel.
 - **Self-hosted, no vendor in the path.** API keys live in `~/.config/mohdel/`. Mohdel calls provider APIs directly; nothing routes through a third party, nothing marks up your tokens, no extra hop of availability risk.
 - **Nothing to compromise.** No network listener, no credential store, no tool execution. Mohdel runs a model call and returns the result; it cannot read a file, run a command, or hand back a key. See [Attack surface](#attack-surface).
 - **Observability without instrumentation.** OpenTelemetry spans, trace-linked logs, and OTLP metrics over one endpoint. Set `OTEL_EXPORTER_OTLP_ENDPOINT`; everything else is wired.
@@ -100,7 +100,7 @@ Model IDs always use the `<provider>/<model>` format:
 
 ```
 openai/gpt-5.6-luna
-anthropic/claude-sonnet-4-6
+anthropic/claude-sonnet-5-5
 openai/gpt-5.4-mini
 groq/llama-4-scout-17b-16e-instruct
 ```
@@ -157,16 +157,16 @@ See [ARCHITECTURE.md §Design principles](ARCHITECTURE.md#design-principles) for
 
 ```bash
 # One-shot inference — pipeable
-mo ask anthropic/claude-sonnet-4-6 "explain monads"
+mo ask anthropic/claude-sonnet-5-5 "explain monads"
 cat article.txt | mo ask openai/gpt-5.4 "summarize in 3 bullets"
 echo "hello" | mo ask openai/gpt-5.6-luna --json | jq .cost
 mo ask openai/gpt-5.6-luna -q "…" 2>err.log   # stderr carries failures only
 
 # Streaming
-mo ask anthropic/claude-sonnet-4-6 --stream "write a haiku about recursion"
+mo ask anthropic/claude-sonnet-5-5 --stream "write a haiku about recursion"
 
 # With thinking effort
-mo ask anthropic/claude-opus-4-6 --effort high "prove P != NP"
+mo ask anthropic/claude-opus-5-5 --effort high "prove P != NP"
 
 # On a faster service lane, when the model sells one
 mo ask openai/gpt-5.6-luna@fast "triage this alert"
@@ -179,7 +179,7 @@ mo transcribe mistral/voxtral-mini-transcribe interview.wav --language fr
 mo ls                                  # list all curated models
 mo ls --sort price                     # sorted by input price
 mo search sonnet                       # filter by name/label
-mo show anthropic/claude-sonnet-4-6    # model details
+mo show anthropic/claude-sonnet-5-5    # model details
 mo stats                               # catalog summary
 mo providers                           # providers with key status & rate limits
 
@@ -201,12 +201,16 @@ mo model instructions anthropic        # brief: fields, doc links, review comman
 mo model check --entry mohdel-candidate.json  # validate + diff, no write
 mo model apply mohdel-candidate.json          # write, after showing the diff
 
+# Copy entries to another host
+mo model export meta/muse-spark-1.3 | ssh host mo model check --entry -     # preview there
+mo model export meta/muse-spark-1.3 | ssh host mo model apply - --yes       # write; remote backup is the undo
+
 # Rate limits
 mo rl show anthropic                   # provider or model limits
-mo rl set anthropic/claude-sonnet-4-6 60 100000
+mo rl set anthropic/claude-sonnet-5-5 60 100000
 
 # Benchmark with live inference
-mo bench anthropic/claude-sonnet-4-6   # single model
+mo bench anthropic/claude-sonnet-5-5   # single model
 mo bench --tag fast --effort low       # suite by tag
 ```
 
@@ -262,6 +266,16 @@ including any field the candidate would remove — before writing. Entries carry
 `source` and `sourcedAt` so a price can be traced back to the page it came
 from. See [docs/CATALOG.md](docs/CATALOG.md#editing-with-a-coding-agent).
 
+### Copying entries to another host
+
+`mo model export <id…>` prints entries exactly as `curated.json` holds them, in the
+shape `mo model apply` reads. `--provider <p>` and `--tag <t>` select in bulk, and
+`--with-redirects` adds the deprecated stubs that point at what you export. Pipe it
+to the other host: `mo model check --entry -` there previews the diff, and
+`mo model apply - --yes` writes it. A pipe has no terminal to confirm on, so the
+remote backup (`mo model backup restore prev`) is the undo. Keys, provider-level
+limits and `catalog.local.json` stay per host.
+
 ## Library Usage
 
 Two integration paths, same adapters underneath: start with the in-process **factory**; graduate to the cross-process **client** when you want gateway-grade isolation.
@@ -272,7 +286,7 @@ Two integration paths, same adapters underneath: start with the in-process **fac
 import mohdel from 'mohdel'
 
 const mo = await mohdel()
-const result = await mo.use('anthropic/claude-sonnet-4-6').answer('Hello')
+const result = await mo.use('anthropic/claude-sonnet-5-5').answer('Hello')
 console.log(result.output, result.cost)
 ```
 

package/config/curated.example.json

@@ -17,18 +17,18 @@
     "tags": ["fast", "chat"]
   },
 
-  "anthropic/claude-sonnet-4-6": {
+  "anthropic/claude-sonnet-5-5": {
     "_comment_b": "Full-featured entry. Adds cache pricing (provider-side prompt caching), thinking effort levels (mohdel translates 'low'/'medium'/'high'/etc. to the provider's native budget), default thinking effort, and a leaderboard tuple. The leaderboard is [intelligence, speed, latency] — used by 'mo rank'.",
-    "model": "claude-sonnet-4-6",
+    "model": "claude-sonnet-5-5",
     "creator": "anthropic",
     "provider": "anthropic",
     "sdk": "anthropic",
-    "label": "Claude Sonnet 4.6",
+    "label": "Claude Sonnet 5.5",
     "inputFormat": ["text", "image"],
-    "inputPrice": 3,
-    "outputPrice": 15,
-    "cacheWritePrice": 3.75,
-    "cacheReadPrice": 0.30,
+    "inputPrice": 2,
+    "outputPrice": 10,
+    "cacheWritePrice": 2.5,
+    "cacheReadPrice": 0.20,
     "contextTokenLimit": 1000000,
     "outputTokenLimit": 128000,
     "defaultThinkingEffort": "medium",
@@ -44,7 +44,7 @@
 
   "anthropic/claude-3-7-sonnet": {
     "_comment_c": "Deprecated stub: a one-field entry that redirects callers to the replacement. 'mo' will refuse to use this id and point at the target. Stubs do NOT need any other fields.",
-    "deprecated": "anthropic/claude-sonnet-4-6"
+    "deprecated": "anthropic/claude-sonnet-5-5"
   },
 
   "novita/flux-2-dev": {

package/config/curated.schema.json

@@ -33,7 +33,7 @@
       "properties": {
         "deprecated": {
           "type": "string",
-          "description": "Replacement model id (e.g. 'anthropic/claude-sonnet-4-6')."
+          "description": "Replacement model id (e.g. 'anthropic/claude-sonnet-5-5')."
         }
       }
     },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mohdel",
-  "version": "3.2.0",
+  "version": "3.4.0",
   "license": "MIT",
   "author": {
     "name": "Christophe Le Bars",
@@ -136,7 +136,7 @@
     "@opentelemetry/exporter-trace-otlp-grpc": "^0.222.0",
     "@opentelemetry/sdk-node": "^0.222.0",
     "chalk": "^6.0.1",
-    "mohdel-thin-gate-linux-x64-gnu": "3.2.0"
+    "mohdel-thin-gate-linux-x64-gnu": "3.4.0"
   },
   "dependencies": {
     "@anthropic-ai/sdk": "^0.129.0",

package/src/cli/ask.js

@@ -82,7 +82,7 @@ Output:
 
 Examples:
   mo ask openai/gpt-5.6-luna "why is the sky blue"
-  cat article.txt | mo ask anthropic/claude-sonnet-4-6 "summarize this"
+  cat article.txt | mo ask anthropic/claude-sonnet-5-5 "summarize this"
   mo ask openai/gpt-5.4 --effort high "explain monads" --json | jq .cost`)
     process.exit(0)
   }

package/src/cli/bench.js

@@ -24,7 +24,7 @@ Options:
   --json                Output as JSON (single model only)
 
 Examples:
-  mo bench anthropic/claude-sonnet-4-6
+  mo bench anthropic/claude-sonnet-5-5
   mo bench --tag fast --effort low
   mo bench openai/gpt-5 --budget 8000 --save results.json`)
     process.exit(0)

package/src/cli/complete.js

@@ -9,7 +9,7 @@ const NOUNS = ['model', 'provider', 'creator', 'tag', 'ratelimit', 'ask', 'trans
 
 const VERBS = {
   model: ['list', 'search', 'stats', 'show', 'get', 'set', 'rm', 'add', 'instructions',
-    'check', 'apply', 'rank', 'bench', 'curate', 'backup'],
+    'check', 'apply', 'export', 'rank', 'bench', 'curate', 'backup'],
   provider: ['list', 'show', 'models', 'setup', 'rm'],
   creator: ['list', 'show'],
   tag: ['list', 'show', 'add', 'rm'],
@@ -24,6 +24,7 @@ const ARGUMENT = {
   'model set': ['model', 'field'],
   'model rm': ['model', 'field'],
   'model bench': ['model'],
+  'model export': ['model'],
   'model curate': ['provider'],
   'model instructions': ['provider'],
   'provider list': ['provider'],

package/src/cli/entry.js

@@ -139,6 +139,7 @@ Write an entry with your coding agent: mo model instructions <provider>`)
   if (!nothingToDo && !args.includes('--yes')) {
     if (!process.stdout.isTTY || path === '-') {
       console.error(`\n${err('Not written')} — no terminal to confirm on. Re-run in a terminal, or pass --yes.`)
+      if (path === '-') console.error(`${meta('Preview without writing:')} mo model check --entry -`)
       process.exit(1)
     }
     const { confirm, isCancel } = await import('@clack/prompts')

package/src/cli/export.js

@@ -0,0 +1,114 @@
+import { catalogEntries } from '../lib/common.js'
+import { loadCuratedCache, getCuratedCacheSnapshot, expandModelAliasSync } from '../lib/curated-cache.js'
+import { providerOf } from '#core/model-id.js'
+import { err } from './colors.js'
+
+const USAGE = 'Usage: mo model export <id…> [--provider <p>] [--tag <t>] [--with-redirects]'
+
+const HELP = `mohdel model export — print catalog entries for another host's \`mo model apply\`
+
+${USAGE}
+
+Prints { "<provider>/<model>": <entry> } as written in curated.json. Ids resolve
+through aliases; one that resolves to nothing fails the command and prints nothing.
+
+  --provider <p>     Every entry of a provider (repeatable)
+  --tag <t>          Every entry carrying a tag (repeatable)
+  --with-redirects   Add the deprecated stubs whose chain ends at an exported entry
+
+Copy to another host — preview, then write (the remote backup is the undo):
+  mo model export meta/muse-spark-1.3 | ssh host mo model check --entry -
+  mo model export meta/muse-spark-1.3 | ssh host mo model apply - --yes`
+
+/**
+ * @param {Record<string, any>} catalog
+ * @param {{ ids?: string[], providers?: string[], tags?: string[], withRedirects?: boolean }} selection
+ * @param {(id: string) => string} [resolve]
+ * @returns {{ entries: Record<string, any>, unknown: string[] }}
+ */
+export const selectEntries = (catalog, { ids = [], providers = [], tags = [], withRedirects = false }, resolve = expandModelAliasSync) => {
+  const keys = new Set()
+  const unknown = []
+
+  for (const requested of ids) {
+    const key = resolve(requested)
+    if (catalog[key]) keys.add(key)
+    else unknown.push(requested)
+  }
+  for (const [key, entry] of catalogEntries(catalog)) {
+    if (providers.includes(providerOf(key))) keys.add(key)
+    if ((entry.tags || []).some(t => tags.includes(t))) keys.add(key)
+  }
+
+  if (withRedirects) {
+    for (const [key, entry] of catalogEntries(catalog)) {
+      if (!entry.deprecated || keys.has(key)) continue
+      const seen = new Set([key])
+      let target = entry.deprecated
+      while (catalog[target]?.deprecated && !seen.has(target)) {
+        seen.add(target)
+        target = catalog[target].deprecated
+      }
+      if (keys.has(target)) {
+        for (const hop of seen) keys.add(hop)
+      }
+    }
+  }
+
+  const entries = {}
+  for (const key of [...keys].sort()) entries[key] = catalog[key]
+  return { entries, unknown }
+}
+
+const usageError = (message) => {
+  console.error(err(message))
+  console.error(USAGE)
+  process.exit(1)
+}
+
+const parseArgs = (args) => {
+  const selection = { ids: [], providers: [], tags: [], withRedirects: false }
+  for (let i = 0; i < args.length; i++) {
+    const arg = args[i]
+    if (arg === '--provider' || arg === '--tag') {
+      const value = args[++i]
+      if (!value || value.startsWith('-')) usageError(`${arg} needs a value`)
+      selection[arg === '--provider' ? 'providers' : 'tags'].push(value)
+    } else if (arg === '--with-redirects') {
+      selection.withRedirects = true
+    } else if (arg.startsWith('-')) {
+      usageError(`Unknown flag: ${arg}`)
+    } else {
+      selection.ids.push(arg)
+    }
+  }
+  return selection
+}
+
+export async function runExport (args) {
+  if (args.includes('-h') || args.includes('--help')) {
+    console.log(HELP)
+    return
+  }
+
+  const { ids, providers, tags, withRedirects } = parseArgs(args)
+
+  if (!ids.length && !providers.length && !tags.length) {
+    console.error(USAGE)
+    process.exit(1)
+  }
+
+  await loadCuratedCache()
+  const { entries, unknown } = selectEntries(getCuratedCacheSnapshot(), { ids, providers, tags, withRedirects })
+
+  if (unknown.length) {
+    console.error(err(`Not in the catalog: ${unknown.join(', ')}`))
+    process.exit(1)
+  }
+  if (!Object.keys(entries).length) {
+    console.error(err('No entry matched.'))
+    process.exit(1)
+  }
+
+  process.stdout.write(`${JSON.stringify(entries, null, 2)}\n`)
+}

package/src/cli/index.js

@@ -40,7 +40,7 @@ if (command === '-h' || command === '--help') {
   const { default: providerDefs } = await import('../lib/providers.js')
   const keyRows = Object.keys(PROVIDER_INFO)
     .map(name => [providerDefs[name].apiKeyEnv, `${PROVIDER_INFO[name].label} API key`])
-    .concat([['MOHDEL_LOCAL_API_SK', 'Bearer token for a local server, if it wants one']])
+    .concat([[providerDefs.local.optionalApiKeyEnv, 'Bearer token for a local server, if it wants one']])
   const keyWidth = Math.max(...keyRows.map(([k]) => k.length))
   const keys = keyRows.map(([k, d]) => `  ${k.padEnd(keyWidth + 2)}${d}`).join('\n')
   console.log(`mohdel — self-hosted LLM gateway and SDK for Node

package/src/cli/model.js

@@ -42,6 +42,7 @@ Usage:
   model instructions [provider]         Print a brief for your coding agent
   model check [--entry <file|->]        Validate catalog or candidate entries
   model apply <file|->                  Write reviewed entries to the catalog
+  model export <id…>                    Print entries for another host's apply
   model backup list|restore|diff        Catalog backups: prev, daily, weekly
   model rank [options]                  Rank models by benchmark performance
   model bench <model> [options]         Benchmark a model with live inference
@@ -377,6 +378,12 @@ config/curated.example.json for ready-to-copy entries.`)
     return
   }
 
+  if (action === 'export') {
+    const { runExport } = await import('./export.js')
+    await runExport(rawArgs.slice(1))
+    return
+  }
+
   if (action === 'instructions') {
     const { runInstructions } = await import('./instructions.js')
     await runInstructions(rawArgs.slice(1))

package/src/lib/providers.js

@@ -1,3 +1,5 @@
+const LOCAL_API_KEY_ENV = 'MOHDEL_LOCAL_API_SK'
+
 // `contextSemantics` and `outputCapStrategy` are published facts about a
 // provider, not switches: mohdel caps `outputBudget` to the model's
 // `outputTokenLimit` whatever they say. They exist so an embedder building its
@@ -82,7 +84,8 @@ const providers = {
     sdk: 'openai',
     api: 'chatCompletions',
     catalog: false,
-    resolveConfiguration: () => ({ apiKey: process.env.MOHDEL_LOCAL_API_SK || '' }),
+    optionalApiKeyEnv: LOCAL_API_KEY_ENV,
+    resolveConfiguration: () => ({ apiKey: process.env[LOCAL_API_KEY_ENV] || '' }),
     contextSemantics: 'shared',
     outputCapStrategy: 'accept'
   },