model-eol 0.2.4 → 0.4.0

package/lib/scanner.mjs

@@ -3,12 +3,13 @@ import path from 'node:path'
 import { spawnSync } from 'node:child_process'
 
 import { buildModelPattern } from './feeds.mjs'
+import { matchesAnyGlob, matchesGlob, normalizeRepoPath } from './glob.mjs'
 
 export const CODE_EXT = new Set([
   '.c', '.cc', '.cjs', '.cpp', '.cs', '.cxx', '.go', '.h', '.hpp', '.java', '.js',
   '.jsx', '.kt', '.kts', '.m', '.mjs', '.mm', '.php', '.py', '.rb', '.rs',
   '.scala', '.sh', '.swift', '.ts', '.tsx',
-  '.cfg', '.env', '.ini', '.json', '.toml', '.yaml', '.yml',
+  '.baml', '.cfg', '.env', '.ini', '.json', '.toml', '.yaml', '.yml',
   '.sql', '.tf', '.tfvars',
 ])
 export const DOC_EXT = new Set(['.md', '.mdx', '.txt', '.rst'])
@@ -33,25 +34,25 @@ export const incompleteScanNotes = notes => notes.filter(note => INCOMPLETE_SCAN
 export const scanIsIncomplete = notes => incompleteScanNotes(notes).length > 0
 
 const directSignals = [
-  ['openai', /\b(OPENAI_API_KEY|api\.openai\.com|from\s+openai\s+import|import\s+OpenAI\s+from\s+['"]openai['"]|require\(['"]openai['"]\)|new\s+OpenAI\s*\()/i, 'OpenAI direct SDK/API'],
-  ['anthropic', /\b(ANTHROPIC_API_KEY|api\.anthropic\.com|@anthropic-ai\/sdk|from\s+anthropic\s+import|import\s+Anthropic\s+from\s+['"]@anthropic-ai\/sdk['"]|new\s+Anthropic\s*\()/i, 'Anthropic direct SDK/API'],
-  ['google-gemini', /\b(GEMINI_API_KEY|GOOGLE_API_KEY|generativelanguage\.googleapis\.com|@google\/genai|google-generativeai|GoogleGenerativeAI)\b/i, 'Gemini direct SDK/API'],
-  ['mistral', /\b(MISTRAL_API_KEY|api\.mistral\.ai|@mistralai\/mistralai|from\s+mistralai\s+import|MistralClient)\b/i, 'Mistral direct SDK/API'],
-  ['cohere', /\b(COHERE_API_KEY|api\.cohere\.ai|from\s+cohere\s+import|new\s+CohereClient\s*\()/i, 'Cohere direct SDK/API'],
-  ['xai', /\b(XAI_API_KEY|api\.x\.ai)\b/i, 'xAI direct SDK/API'],
+  ['openai', /\b(OPENAI_API_KEY|api\.openai\.com|from\s+openai\s+import|import\s+OpenAI\s+from\s+['"]openai['"]|require\(['"]openai['"]\)|new\s+OpenAI\s*\(|provider\s+['"]?openai\b)/i, 'OpenAI direct SDK/API or BAML provider'],
+  ['anthropic', /\b(ANTHROPIC_API_KEY|api\.anthropic\.com|@anthropic-ai\/sdk|from\s+anthropic\s+import|import\s+Anthropic\s+from\s+['"]@anthropic-ai\/sdk['"]|new\s+Anthropic\s*\(|provider\s+['"]?anthropic\b)/i, 'Anthropic direct SDK/API or BAML provider'],
+  ['google-gemini', /\b(GEMINI_API_KEY|GOOGLE_API_KEY|generativelanguage\.googleapis\.com|@google\/genai|google-generativeai|GoogleGenerativeAI|provider\s+['"]?google-ai\b)\b/i, 'Gemini direct SDK/API or BAML provider'],
+  ['mistral', /\b(MISTRAL_API_KEY|api\.mistral\.ai|@mistralai\/mistralai|from\s+mistralai\s+import|MistralClient|provider\s+['"]?mistral\b)\b/i, 'Mistral direct SDK/API or BAML provider'],
+  ['cohere', /\b(COHERE_API_KEY|api\.cohere\.ai|from\s+cohere\s+import|new\s+CohereClient\s*\(|provider\s+['"]?cohere\b)/i, 'Cohere direct SDK/API or BAML provider'],
+  ['xai', /\b(XAI_API_KEY|api\.x\.ai|provider\s+['"]?xai\b)\b/i, 'xAI direct SDK/API or BAML provider'],
 ]
 
 const cloudSignals = [
-  ['azure-ai-foundry', /\b(AZURE_OPENAI|AzureOpenAI|@azure\/openai|azure\.ai\.openai|azure-ai-foundry|AZURE_AI)\b/i, 'Azure Foundry/OpenAI deployment reference'],
-  ['aws-bedrock', /\b(BEDROCK_MODEL_ID|bedrock-runtime|BedrockRuntime|InvokeModel|invoke_model|amazon\.bedrock)\b/i, 'Amazon Bedrock reference'],
-  ['vertex-ai', /\b(VERTEX_AI|GOOGLE_CLOUD_PROJECT|aiplatform|vertexai|VertexAI|PredictionServiceClient|publishers\/google\/models)\b/i, 'Vertex AI reference'],
+  ['azure-ai-foundry', /\b(AZURE_OPENAI|AzureOpenAI|@azure\/openai|azure\.ai\.openai|azure-ai-foundry|AZURE_AI|provider\s+['"]?azure-openai\b)\b/i, 'Azure Foundry/OpenAI deployment reference'],
+  ['aws-bedrock', /\b(BEDROCK_MODEL_ID|bedrock-runtime|BedrockRuntime|InvokeModel|invoke_model|amazon\.bedrock|provider\s+['"]?aws-bedrock\b)\b/i, 'Amazon Bedrock reference'],
+  ['vertex-ai', /\b(VERTEX_AI|GOOGLE_CLOUD_PROJECT|aiplatform|vertexai|VertexAI|PredictionServiceClient|publishers\/google\/models|provider\s+['"]?vertex-ai\b)\b/i, 'Vertex AI reference'],
 ]
 
 const gatewaySignals = [
-  ['openrouter', /\b(OPENROUTER_API_KEY|openrouter\.ai|openrouter\/)\b/i, 'OpenRouter gateway reference'],
-  ['litellm', /\b(LITELLM|litellm|LiteLLM)\b/, 'LiteLLM gateway reference'],
-  ['portkey', /\b(PORTKEY_API_KEY|portkey|Portkey)\b/, 'Portkey gateway reference'],
-  ['ai-gateway', /\b(AI_GATEWAY|Vercel AI Gateway|ai-gateway)\b/i, 'AI gateway reference'],
+  ['openrouter', /\b(OPENROUTER_API_KEY|openrouter\.ai|openrouter\/|provider\s+['"]?openrouter\b)\b/i, 'OpenRouter gateway reference'],
+  ['litellm', /\b(LITELLM|litellm|LiteLLM|provider\s+['"]?litellm\b)\b/, 'LiteLLM gateway reference'],
+  ['portkey', /\b(PORTKEY_API_KEY|portkey|Portkey|provider\s+['"]?portkey\b)\b/, 'Portkey gateway reference'],
+  ['ai-gateway', /\b(AI_GATEWAY|Vercel AI Gateway|ai-gateway|provider\s+['"]?ai-gateway\b)\b/i, 'AI gateway reference'],
 ]
 
 const allSignals = [
@@ -71,6 +72,13 @@ const candidatePatterns = [
   /["']((?:gpt|o[1345]|claude|gemini|mistral|command|grok|llama|deepseek|qwen|nova|embed)[A-Za-z0-9._:/-]{4,120})["']/gi,
 ]
 
+const obviousCandidateNoise = value =>
+  /^[A-Z][A-Z0-9_]+$/.test(value) ||
+  /^commands?_[a-z0-9_]+$/i.test(value) ||
+  /^claude_code[._]/i.test(value) ||
+  /\.(?:md|mdx|json|ya?ml|toml|txt|rst)$/i.test(value) ||
+  /\.(?:token|cost)\.usage$/i.test(value)
+
 const suffixes = value => {
   const parts = value.split(/[\/.:]/).filter(Boolean)
   return parts.flatMap((_, i) => [parts.slice(i).join('/'), parts.slice(i).join('.'), parts.slice(i).join(':')])
@@ -87,7 +95,7 @@ const findCandidates = (lineText, entries) => {
     let m
     while ((m = pattern.exec(lineText)) !== null) {
       const value = m[1].trim()
-      if (seen.has(value) || isKnownModel(value, entries)) continue
+      if (seen.has(value)) continue
       seen.add(value)
       candidates.push(value)
     }
@@ -95,6 +103,10 @@ const findCandidates = (lineText, entries) => {
   return candidates
 }
 
+const publisherForProvider = provider => provider === 'google-gemini' ? 'google' : provider
+
+const providerMatchesPublisher = (provider, publisher) => publisherForProvider(provider) === publisher
+
 const classifyRef = ({ file, line, lines, publisher }) => {
   const start = Math.max(0, line - 8)
   const end = Math.min(lines.length, line + 4)
@@ -108,8 +120,12 @@ const classifyRef = ({ file, line, lines, publisher }) => {
   const nearest = signals[0]
   if (nearest) {
     const hasCompetingRoute = signals.some(signal => signal.usage === 'cloud-provider' || signal.usage === 'gateway')
-    const confidence = nearest.usage === 'direct-api' && !hasCompetingRoute ? 'high' : 'medium'
-    return { usage: nearest.usage, provider: nearest.provider, confidence, evidence: nearest.evidence }
+    const providerMatches = nearest.usage !== 'direct-api' || providerMatchesPublisher(nearest.provider, publisher)
+    const confidence = nearest.usage === 'direct-api' && !hasCompetingRoute && providerMatches ? 'high' : 'medium'
+    const evidence = providerMatches
+      ? nearest.evidence
+      : `${nearest.evidence}; signal provider ${nearest.provider} differs from model publisher ${publisher}`
+    return { usage: nearest.usage, provider: nearest.provider, confidence, evidence }
   }
 
   const lowerFile = file.toLowerCase()
@@ -122,6 +138,7 @@ const classifyRef = ({ file, line, lines, publisher }) => {
 
 const isScannable = (file, includeDocs) => {
   const basename = path.basename(file)
+  if (basename === '.model-eol.json') return false
   if (basename === '.env' || basename.startsWith('.env.')) return true
   const ext = path.extname(file)
   return CODE_EXT.has(ext) || (includeDocs && DOC_EXT.has(ext))
@@ -241,17 +258,66 @@ export const addedLinesForTargets = (targets, baseRef) => {
 export const filterFindingsToChanged = (findings, addedLines) =>
   findings.filter(finding => addedLines.has(changedKey(finding.file, finding.line)))
 
-const collectFilesDetailed = (targets, { includeDocs = false } = {}) => {
+const collectFilesDetailed = (targets, {
+  includeDocs = false,
+  ignorePaths = [],
+  ignoreRoots = [],
+  ignoredFiles = [],
+  policyForFile = null,
+} = {}) => {
   const files = []
   const notes = []
   const validTargets = []
   const seen = new Set()
+  const canonicalPath = value => {
+    const resolved = path.resolve(value)
+    try {
+      return fs.realpathSync(resolved)
+    } catch {
+      return resolved
+    }
+  }
+  const excludedFiles = new Set(ignoredFiles.map(canonicalPath))
+  const roots = ignoreRoots.map(canonicalPath)
   let fileLimitWarned = false
   let stopped = false
 
   const note = value => notes.push(value)
+  const repoPathsFor = absolute => {
+    const resolved = canonicalPath(absolute)
+    const repoPaths = []
+    for (const root of roots) {
+      const relative = path.relative(root, resolved)
+      if (relative === '' || relative === '..' || relative.startsWith(`..${path.sep}`) || path.isAbsolute(relative)) continue
+      repoPaths.push(normalizeRepoPath(relative))
+    }
+    return repoPaths
+  }
+  const matchesIgnoredRepoPath = repoPath => matchesAnyGlob(repoPath, ignorePaths) ||
+    ignorePaths.some(pattern => {
+      const normalizedPattern = normalizeRepoPath(pattern).replace(/\/+$/, '')
+      return normalizedPattern && !/[*?]/.test(normalizedPattern) && repoPath.startsWith(`${normalizedPattern}/`)
+    })
+  const isIgnoredFile = absolute => {
+    const resolved = canonicalPath(absolute)
+    if (excludedFiles.has(resolved) || repoPathsFor(resolved).some(matchesIgnoredRepoPath)) return true
+    const policy = policyForFile?.(resolved)
+    return Boolean(policy?.repoPath && matchesIgnoredRepoPathWithPatterns(policy.repoPath, policy.ignore?.paths ?? []))
+  }
+  const matchesIgnoredRepoPathWithPatterns = (repoPath, patterns) => matchesAnyGlob(repoPath, patterns) ||
+    patterns.some(pattern => {
+      const normalizedPattern = normalizeRepoPath(pattern).replace(/\/+$/, '')
+      return normalizedPattern && !/[*?]/.test(normalizedPattern) && repoPath.startsWith(`${normalizedPattern}/`)
+    })
+  const isIgnoredDirectoryTree = absolute => repoPathsFor(absolute).some(repoPath =>
+    ignorePaths.some(pattern => {
+      const normalizedPattern = normalizeRepoPath(pattern)
+      const literalDirectory = !/[*?]/.test(normalizedPattern) && normalizedPattern.replace(/\/+$/, '') === repoPath
+      const subtreeGlob = normalizedPattern.endsWith('/**') && matchesGlob(repoPath, normalizedPattern.slice(0, -3))
+      return literalDirectory || subtreeGlob
+    }))
   const addFile = absolute => {
-    if (stopped || seen.has(absolute)) return
+    if (stopped || seen.has(absolute) || isIgnoredFile(absolute)) return
     seen.add(absolute)
 
     let st
@@ -288,7 +354,7 @@ const collectFilesDetailed = (targets, { includeDocs = false } = {}) => {
   }
 
   const walk = absolute => {
-    if (stopped) return
+    if (stopped || excludedFiles.has(canonicalPath(absolute))) return
     let st
     try {
       st = fs.lstatSync(absolute)
@@ -301,6 +367,7 @@ const collectFilesDetailed = (targets, { includeDocs = false } = {}) => {
       addFile(absolute)
       return
     }
+    if (isIgnoredDirectoryTree(absolute)) return
     if (SKIP_DIRS.has(path.basename(absolute))) return
     let names
     try {
@@ -373,25 +440,101 @@ const collectFilesDetailed = (targets, { includeDocs = false } = {}) => {
 
 export const collectFiles = (targets, options = {}) => collectFilesDetailed(targets, options).files
 
-export const scanTargets = ({ targets, entries, keys, includeDocs = false }) => {
-  const collected = collectFilesDetailed(targets, { includeDocs })
+export const scanTargets = ({
+  targets,
+  entries,
+  keys,
+  includeDocs = false,
+  ignoreModels = [],
+  ignorePaths = [],
+  ignoreRoots = [],
+  ignoredFiles = [],
+  policyForFile = null,
+  routeForReference = null,
+  mappedRoutesForFile = null,
+}) => {
+  const collected = collectFilesDetailed(targets, { includeDocs, ignorePaths, ignoreRoots, ignoredFiles, policyForFile })
   if (collected.validTargets.length === 0) throw new Error('no valid targets remain')
   const { files } = collected
   const pattern = buildModelPattern(keys)
+  const ignoredSets = models => ({
+    keys: new Set(models),
+    canonicalIds: new Set(models.map(key => entries.get(key)?.entry.id ?? key)),
+  })
+  const baseIgnored = ignoredSets(ignoreModels)
+  const ignoresModel = (file, matched, id = null) => {
+    const configured = policyForFile?.(path.resolve(file))?.ignore?.models ?? []
+    const dynamicIgnored = configured.length ? ignoredSets(configured) : null
+    return baseIgnored.keys.has(matched) || (id !== null && baseIgnored.canonicalIds.has(id)) ||
+      dynamicIgnored?.keys.has(matched) || (id !== null && dynamicIgnored?.canonicalIds.has(id))
+  }
   const modelRefs = []
   const candidateRefs = []
   const integrationHints = []
 
-  const isFeedDocument = (file, text) => {
+  const modelEolDocument = (file, text) => {
     if (!file.endsWith('.json')) return false
     try {
-      const spec = JSON.parse(text)?.spec
-      return typeof spec === 'string' && spec.startsWith('model-eol/')
+      const value = JSON.parse(text)
+      const productSchema = [value?.spec, value?.schema, value?.plan_schema]
+        .some(identifier => typeof identifier === 'string' && (identifier.startsWith('model-eol/') || identifier.startsWith('model-eol.')))
+      const generatedCycloneDx = value?.bomFormat === 'CycloneDX' &&
+        value?.metadata?.properties?.some(property =>
+          property?.name === 'model-eol:generator' && property?.value === 'model-eol/inventory-cyclonedx@0.1')
+      return productSchema || generatedCycloneDx
     } catch {
       return false
     }
   }
 
+  const generatedSource = text => {
+    const header = text.slice(0, 8192)
+    return /This file was generated by BAML/i.test(header) ||
+      /Code generated .* DO NOT EDIT/i.test(header) ||
+      /^\s*(?:[/#*;-]+\s*)?@generated\b/im.test(header) ||
+      /AUTO[- ]GENERATED FILE[\s\S]{0,240}DO NOT EDIT/i.test(header)
+  }
+
+  const routeClassification = route => {
+    const usage = cloudSignals.some(([provider]) => provider === route.via)
+      ? 'cloud-provider'
+      : gatewaySignals.some(([provider]) => provider === route.via)
+        ? 'gateway'
+        : 'model-reference'
+    return {
+      usage,
+      provider: route.via,
+      confidence: 'high',
+      evidence: `configured route ${route.index} via ${route.via}`,
+    }
+  }
+
+  const pushModelRef = ({ file, line, lineText, matchIndex = null, matched, record, classification, route = null, mappedFrom = null }) => {
+    if (ignoresModel(file, matched, record.entry.id) || (mappedFrom && ignoresModel(file, mappedFrom, record.entry.id))) return
+    const occurrence = [...lineText.matchAll(new RegExp(escapeRegExp(matched), 'g'))]
+      .findIndex(match => match.index === (matchIndex ?? lineText.indexOf(matched)))
+    modelRefs.push({
+      kind: 'model-reference',
+      file,
+      line,
+      matched,
+      id: record.entry.id,
+      entry: record.entry,
+      publisher: record.publisher,
+      source: record.source,
+      feedNote: record.feedNote,
+      policy: record.policy,
+      generated: record.generated,
+      usage: classification.usage,
+      resolved_provider: classification.provider,
+      confidence: classification.confidence,
+      evidence: classification.evidence,
+      occurrence: occurrence < 0 ? 0 : occurrence,
+      route_index: route?.index ?? null,
+      mapped_from: mappedFrom,
+    })
+  }
+
   for (const file of files) {
     let text
     try {
@@ -400,59 +543,81 @@ export const scanTargets = ({ targets, entries, keys, includeDocs = false }) =>
       collected.notes.push({ reason: 'unreadable-file', file, message: e.message })
       continue
     }
-    // A feed document is lifecycle data, not model usage.
-    if (isFeedDocument(file, text)) {
-      collected.notes.push({ reason: 'feed-document-skipped', file })
+    // Product artifacts and generated clients describe model IDs but are not
+    // authoritative usage or safe migration targets.
+    if (modelEolDocument(file, text)) {
+      collected.notes.push({ reason: 'model-eol-document-skipped', file })
+      continue
+    }
+    if (generatedSource(text)) {
+      collected.notes.push({ reason: 'generated-artifact-skipped', file })
       continue
     }
     const lines = text.split('\n')
     for (const [i, lineText] of lines.entries()) {
-      for (const signal of findSignals(lineText)) {
-        integrationHints.push({
-          kind: 'integration-hint',
-          usage: signal.usage,
-          provider: signal.provider,
-          file,
-          line: i + 1,
-          matched: signal.matched,
-          evidence: signal.evidence,
-        })
+      const signals = findSignals(lineText)
+      const candidates = findCandidates(lineText, entries)
+      const mappedSources = new Set()
+      const mappedSpans = []
+      const resolvedProviders = new Set()
+      const selectedMappings = new Map()
+      for (const configuredRoute of mappedRoutesForFile?.(file) ?? []) {
+        const route = routeForReference?.({ file, matched: configuredRoute.match, id: null }) ?? null
+        if (!route?.model) continue
+        selectedMappings.set(`${route.index}:${route.match}`, route)
+      }
+      for (const route of selectedMappings.values()) {
+        const record = entries.get(route.model)
+        if (!record) continue
+        const exactRoutePattern = new RegExp(`(?<![A-Za-z0-9._:/-])${escapeRegExp(route.match)}(?![A-Za-z0-9._:/-])`, 'g')
+        for (const match of lineText.matchAll(exactRoutePattern)) {
+          mappedSources.add(route.match)
+          mappedSpans.push({ start: match.index, end: match.index + route.match.length })
+          resolvedProviders.add(route.via)
+          pushModelRef({
+            file,
+            line: i + 1,
+            lineText,
+            matchIndex: match.index,
+            matched: route.match,
+            record,
+            classification: routeClassification(route),
+            route,
+            mappedFrom: route.match,
+          })
+        }
       }
 
-      if (!pattern) continue
-      pattern.lastIndex = 0
-      let m
-      while ((m = pattern.exec(lineText)) !== null) {
-        const record = entries.get(m[1])
-        const occurrence = [...lineText.matchAll(new RegExp(escapeRegExp(m[1]), 'g'))]
-          .findIndex(match => match.index === m.index)
-        const classification = classifyRef({
-          file,
-          line: i + 1,
-          lines,
-          publisher: record.publisher,
-        })
-        modelRefs.push({
-          kind: 'model-reference',
-          file,
-          line: i + 1,
-          matched: m[1],
-          id: record.entry.id,
-          entry: record.entry,
-          publisher: record.publisher,
-          source: record.source,
-          feedNote: record.feedNote,
-          policy: record.policy,
-          generated: record.generated,
-          usage: classification.usage,
-          resolved_provider: classification.provider,
-          confidence: classification.confidence,
-          evidence: classification.evidence,
-          occurrence: occurrence < 0 ? 0 : occurrence,
-        })
+      if (pattern) {
+        pattern.lastIndex = 0
+        let m
+        while ((m = pattern.exec(lineText)) !== null) {
+          if (mappedSpans.some(span => m.index >= span.start && m.index + m[1].length <= span.end)) continue
+          const record = entries.get(m[1])
+          const route = routeForReference?.({ file, matched: m[1], id: record.entry.id }) ?? null
+          const classification = route ? routeClassification(route) : classifyRef({
+            file,
+            line: i + 1,
+            lines,
+            publisher: record.publisher,
+          })
+          if (route) resolvedProviders.add(route.via)
+          pushModelRef({
+            file,
+            line: i + 1,
+            lineText,
+            matchIndex: m.index,
+            matched: m[1],
+            record,
+            classification,
+            route,
+          })
+        }
       }
 
-      for (const candidate of findCandidates(lineText, entries)) {
+      for (const candidate of candidates) {
+        if (mappedSources.has(candidate) || isKnownModel(candidate, entries) || obviousCandidateNoise(candidate)) continue
+        if (ignoresModel(file, candidate)) continue
         const classification = classifyRef({
           file,
           line: i + 1,
@@ -471,6 +636,18 @@ export const scanTargets = ({ targets, entries, keys, includeDocs = false }) =>
           reason: 'model-like string not present in loaded feeds',
         })
       }
+      for (const signal of signals) {
+        if (resolvedProviders.has(signal.provider)) continue
+        integrationHints.push({
+          kind: 'integration-hint',
+          usage: signal.usage,
+          provider: signal.provider,
+          file,
+          line: i + 1,
+          matched: signal.matched,
+          evidence: signal.evidence,
+        })
+      }
     }
   }
 

package/package.json

@@ -1,13 +1,14 @@
 {
   "name": "model-eol",
-  "version": "0.2.4",
-  "description": "A machine-readable deprecation feed format for AI models, plus a zero-dependency checker and inventory CLI.",
+  "version": "0.4.0",
+  "description": "AI model deprecation feeds with a zero-dependency checker, inventory CLI, and migration bot.",
   "type": "module",
   "bin": {
-    "model-eol": "check.mjs"
+    "model-eol": "check.mjs",
+    "model-eol-bot": "bot/bot.mjs"
   },
   "scripts": {
-    "test": "node test/run.mjs && node refresh/test/run.mjs && node bot/test/run.mjs && node scripts/test-feed-changelog.mjs"
+    "test": "node test/run.mjs && node refresh/test/run.mjs && node bot/test/run.mjs && node scripts/test-action-contract.mjs && node bot/test/package.mjs && node scripts/test-feed-changelog.mjs"
   },
   "repository": {
     "type": "git",
@@ -31,7 +32,10 @@
   ],
   "files": [
     "check.mjs",
+    "bot.yml.example",
     "lib/",
+    "bot/bot.mjs",
+    "bot/lib/",
     "feeds/",
     "schema/",
     "SPEC.md",

package/schema/model-eol.bot-config.schema.json

@@ -1,8 +1,8 @@
 {
   "$schema": "http://json-schema.org/draft-07/schema#",
   "$id": "https://example.invalid/model-eol/0.1/bot-config.schema.json",
-  "title": "model-eol bot configuration",
-  "description": "Optional .model-eol.json configuration for the zero-dependency GitHub bot adapter.",
+  "title": "model-eol repository configuration",
+  "description": "Optional strict .model-eol.json configuration shared by the zero-dependency CLI, Action, and GitHub bot adapter.",
   "type": "object",
   "additionalProperties": false,
   "properties": {
@@ -15,10 +15,11 @@
     "scope": {
       "enum": ["direct", "all"],
       "default": "direct",
-      "description": "Use direct API references only, or include generic model references."
+      "description": "Use the direct-first scope (direct API plus generic model references, leaving cloud/gateway references as hints), or include all tracked references."
     },
     "via": {
       "type": ["string", "null"],
+      "minLength": 1,
       "default": null,
       "description": "Optional distributor lifecycle clock, such as azure-ai-foundry."
     },
@@ -42,17 +43,29 @@
           "type": "array",
           "items": { "type": "string", "minLength": 1 },
           "default": [],
-          "description": "Canonical IDs or aliases to ignore. A canonical ID ignores all of its aliases."
+          "description": "Canonical IDs or aliases to ignore across CLI, Action, and bot. Either form ignores every alias of the canonical feed entry."
         },
         "paths": {
           "type": "array",
           "items": { "type": "string", "minLength": 1 },
           "default": [],
-          "description": "Repo-relative simple globs supporting *, **, and ?."
+          "description": "Repo-relative simple globs supporting *, **, and ?. Matching paths are excluded before file coverage checks where possible."
         }
       },
       "default": { "models": [], "paths": [] }
     },
+    "overrides": {
+      "type": "array",
+      "items": { "$ref": "#/definitions/pathOverride" },
+      "default": [],
+      "description": "Path-scoped policy overrides. Matching entries apply in array order; later scalar values win and ignore lists are additive."
+    },
+    "routes": {
+      "type": "array",
+      "items": { "$ref": "#/definitions/referenceRoute" },
+      "default": [],
+      "description": "Path- and optionally reference-scoped distributor routes. The last matching route wins. Exact match/model pairs can resolve repository deployment aliases to canonical feed IDs."
+    },
     "issues": {
       "type": "object",
       "additionalProperties": false,
@@ -71,17 +84,20 @@
       "properties": {
         "command": {
           "type": ["string", "null"],
+          "minLength": 1,
           "default": null,
-          "description": "Opt-in command run in the temporary clone for changed patches."
+          "description": "Opt-in evaluation command. Direct bot --eval runs it per changed patch in a temporary clone; the split example workflow runs it once as a read-only plan preflight."
         },
         "timeout_ms": {
           "type": "integer",
-          "minimum": 1,
+          "minimum": 1000,
+          "maximum": 3600000,
           "default": 600000
         },
         "max_report_bytes": {
           "type": "integer",
-          "minimum": 0,
+          "minimum": 1,
+          "maximum": 8388608,
           "default": 65536
         },
         "pass_env": {
@@ -100,7 +116,56 @@
     "via": null,
     "feeds": { "allow_vendored_fallback": false },
     "ignore": { "models": [], "paths": [] },
+    "overrides": [],
+    "routes": [],
     "issues": { "enabled": true },
     "eval": { "command": null, "timeout_ms": 600000, "max_report_bytes": 65536, "pass_env": [] }
+  },
+  "definitions": {
+    "stringArray": {
+      "type": "array",
+      "items": { "type": "string", "minLength": 1 }
+    },
+    "nonEmptyStringArray": {
+      "allOf": [
+        { "$ref": "#/definitions/stringArray" },
+        { "minItems": 1 }
+      ]
+    },
+    "ignore": {
+      "type": "object",
+      "additionalProperties": false,
+      "properties": {
+        "models": { "$ref": "#/definitions/stringArray" },
+        "paths": { "$ref": "#/definitions/stringArray" }
+      }
+    },
+    "pathOverride": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["paths"],
+      "properties": {
+        "paths": { "$ref": "#/definitions/nonEmptyStringArray" },
+        "days": { "type": "integer", "minimum": 0 },
+        "scope": { "enum": ["direct", "all"] },
+        "via": { "type": ["string", "null"], "minLength": 1 },
+        "ignore": { "$ref": "#/definitions/ignore" }
+      }
+    },
+    "referenceRoute": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["paths", "via"],
+      "properties": {
+        "paths": { "$ref": "#/definitions/nonEmptyStringArray" },
+        "via": { "type": "string", "minLength": 1 },
+        "match": { "type": "string", "minLength": 1 },
+        "model": { "type": "string", "minLength": 1 }
+      },
+      "dependencies": {
+        "match": ["model"],
+        "model": ["match"]
+      }
+    }
   }
 }

package/schema/model-eol.inventory.schema.json

@@ -60,8 +60,15 @@
             "evidence": { "type": "string" },
             "status": { "$ref": "#/definitions/status" },
             "shutdown": { "type": ["string", "null"] },
+            "date_precision": { "enum": ["exact", "earliest", null] },
+            "distribution_status": { "enum": ["active", "legacy", "extended-access", "retired", null] },
             "via": { "type": ["string", "null"] },
+            "requested_via": { "type": ["string", "null"] },
             "days": { "type": ["integer", "null"] },
+            "threshold_days": { "type": "integer", "minimum": 0 },
+            "effective_scope": { "enum": ["all", "direct"] },
+            "policy_provenance": { "$ref": "#/definitions/policyProvenance" },
+            "mapped_from": { "type": "string", "minLength": 1 },
             "safe_until": { "type": ["string", "null"], "pattern": "^\\d{4}-\\d{2}-\\d{2}$" },
             "replacement": { "type": ["string", "null"] },
             "source": { "type": ["string", "null"] },
@@ -92,6 +99,8 @@
             "resolved_provider": { "type": "string" },
             "confidence": { "$ref": "#/definitions/confidence" },
             "evidence": { "type": "string" },
+            "effective_scope": { "enum": ["all", "direct"] },
+            "policy_provenance": { "$ref": "#/definitions/policyProvenance" },
             "reason": { "type": "string" }
           }
         }
@@ -125,6 +134,18 @@
         "limit_files": { "type": "integer", "minimum": 0 }
       },
       "additionalProperties": false
+    },
+    "policyProvenance": {
+      "type": "object",
+      "required": ["override_indexes", "route_index"],
+      "properties": {
+        "override_indexes": {
+          "type": "array",
+          "items": { "type": "integer", "minimum": 0 }
+        },
+        "route_index": { "type": ["integer", "null"], "minimum": 0 }
+      },
+      "additionalProperties": false
     }
   }
 }