pi-requesty-discover 1.0.6 → 1.2.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-requesty-discover",
-  "version": "1.0.6",
+  "version": "1.2.0",
   "description": "Discover available Requesty models for the Pi Coding Agent",
   "repository": {
     "type": "git",

package/src/health-check.ts

@@ -1,8 +1,8 @@
 import type { ProviderModelConfig } from '@earendil-works/pi-coding-agent'
 import fs from 'node:fs'
 import path from 'node:path'
-import { z } from 'zod'
 import { getEnv, type Env } from './env'
+import { formatModelsDiffSummary, type ModelsDiff } from './models-json'
 
 const HEALTH_CHECK_CONCURRENCY = 10
 const HEALTH_CHECK_TIMEOUT_MS = 15_000
@@ -16,9 +16,7 @@ const defaultHealthCheckOptions = {
   retryDelayMs: HEALTH_CHECK_RETRY_DELAY_MS,
 }
 
-const ChatCompletionResponseSchema = z.object({
-  choices: z.array(z.unknown()).min(1),
-})
+const DATA_PREFIX = 'data:'
 
 export type Provider = {
   baseUrl: string
@@ -101,7 +99,12 @@ export function formatHealthSummary(results: HealthCheckResult[]): string {
   return `Health check: ${passed.length} OK, ${failed.length} failed:\n${failedModels}\n`
 }
 
-export function writeHealthCheckLog(provider: Provider, results: HealthCheckResult[], envConfig: Env = getEnv()): void {
+export function writeHealthCheckLog(
+  provider: Provider,
+  results: HealthCheckResult[],
+  diff: ModelsDiff,
+  envConfig: Env = getEnv(),
+): void {
   const passed = results.filter(r => r.ok)
   const failed = results.filter(r => !r.ok)
   const lines = [
@@ -113,6 +116,8 @@ export function writeHealthCheckLog(provider: Provider, results: HealthCheckResu
     `Passed: ${passed.length}`,
     `Failed: ${failed.length}`,
     '',
+    formatModelsDiffSummary(diff),
+    '',
   ]
 
   if (failed.length === 0) {
@@ -138,48 +143,92 @@ export function writeHealthCheckLog(provider: Provider, results: HealthCheckResu
 
 export async function postChatCompletion(
   provider: Provider,
-  body: unknown,
+  body: Record<string, unknown>,
   options: HealthCheckOptions = {},
 ): Promise<ModelCheckResult> {
   const healthCheckOptions = resolveHealthCheckOptions(options)
   const start = Date.now()
-  try {
-    const response = await fetchWithTimeoutRetries(
-      `${provider.baseUrl}/chat/completions`,
-      {
+
+  for (let attempt = 0; attempt <= healthCheckOptions.retries; attempt++) {
+    try {
+      const response = await fetch(`${provider.baseUrl}/chat/completions`, {
         method: 'POST',
         headers: {
           Authorization: `Bearer ${provider.apiKey}`,
           'Content-Type': 'application/json',
+          Accept: 'text/event-stream',
         },
-        body: JSON.stringify(body),
-      },
-      healthCheckOptions,
-    )
+        body: JSON.stringify({ ...body, stream: true }),
+        signal: AbortSignal.timeout(healthCheckOptions.timeoutMs),
+      })
 
-    const latencyMs = Date.now() - start
+      const latencyMs = Date.now() - start
 
-    if (!response.ok) {
-      const text = await response.text().catch(() => '')
-      return { ok: false, latencyMs, error: `HTTP ${response.status} ${response.statusText}${text ? `: ${text}` : ''}` }
-    }
+      if (!response.ok) {
+        const text = await response.text().catch(() => '')
+        return {
+          ok: false,
+          latencyMs,
+          error: `HTTP ${response.status} ${response.statusText}${text ? `: ${text}` : ''}`,
+        }
+      }
 
-    const result = ChatCompletionResponseSchema.safeParse(await response.json())
-    if (!result.success) {
-      return { ok: false, latencyMs, error: 'Empty choices in response' }
+      return await verifyFirstStreamChunk(response.body!, start)
+    } catch (err) {
+      if (isTimeoutError(err) && attempt < healthCheckOptions.retries) {
+        if (healthCheckOptions.retryDelayMs > 0) {
+          await new Promise(resolve => setTimeout(resolve, healthCheckOptions.retryDelayMs))
+        }
+        continue
+      }
+      const attempts = attempt + 1
+      return {
+        ok: false,
+        latencyMs: Date.now() - start,
+        error: isTimeoutError(err)
+          ? `Timed out after ${attempts} attempt(s); per-attempt timeout is ${healthCheckOptions.timeoutMs / 1000}s`
+          : String(err),
+      }
     }
+  }
 
-    return { ok: true, latencyMs }
-  } catch (err) {
-    const latencyMs = Date.now() - start
-    const attempts = healthCheckOptions.retries + 1
-    return {
-      ok: false,
-      latencyMs,
-      error: isTimeoutError(err)
-        ? `Timed out after ${attempts} attempt(s); per-attempt timeout is ${healthCheckOptions.timeoutMs / 1000}s`
-        : String(err),
+  return { ok: false, latencyMs: Date.now() - start, error: 'Unknown error' }
+}
+
+async function verifyFirstStreamChunk(body: ReadableStream<Uint8Array>, start: number): Promise<ModelCheckResult> {
+  const reader = body.getReader()
+  const decoder = new TextDecoder()
+  let tail = ''
+  try {
+    while (true) {
+      const { done, value } = await reader.read()
+      if (done) {
+        break
+      }
+      tail += decoder.decode(value, { stream: true })
+      const lines = tail.split('\n')
+      tail = lines.pop()!
+      for (const line of lines) {
+        if (!line.startsWith(DATA_PREFIX)) {
+          continue
+        }
+        const payload = line.slice(DATA_PREFIX.length).trim()
+        if (payload === '[DONE]') {
+          continue
+        }
+        try {
+          const parsed = JSON.parse(payload) as { choices?: unknown }
+          if (Array.isArray(parsed.choices) && parsed.choices.length > 0) {
+            return { ok: true, latencyMs: Date.now() - start }
+          }
+        } catch {
+          // partial JSON or non-choices chunk — keep reading
+        }
+      }
     }
+    return { ok: false, latencyMs: Date.now() - start, error: 'Stream ended without content' }
+  } finally {
+    await reader.cancel()
   }
 }
 
@@ -248,34 +297,6 @@ function resolveHealthCheckOptions(options: HealthCheckOptions): ResolvedHealthC
   }
 }
 
-async function fetchWithTimeoutRetries(
-  url: string,
-  options: RequestInit,
-  { timeoutMs, retries, retryDelayMs }: ResolvedHealthCheckOptions,
-) {
-  function sleep(ms: number) {
-    if (ms <= 0) {
-      return
-    }
-    return new Promise(resolve => setTimeout(resolve, ms))
-  }
-
-  for (let attempt = 0; ; attempt++) {
-    try {
-      return await fetch(url, {
-        ...options,
-        signal: AbortSignal.timeout(timeoutMs),
-      })
-    } catch (error) {
-      if (!isTimeoutError(error) || attempt >= retries) {
-        throw error
-      }
-
-      await sleep(retryDelayMs)
-    }
-  }
-}
-
 function isTimeoutError(error: unknown): boolean {
   return error instanceof Error && (error.name === 'AbortError' || error.name === 'TimeoutError')
 }

package/src/index.ts

@@ -1,6 +1,12 @@
 import { ExtensionAPI, ExtensionCommandContext, ProviderModelConfig } from '@earendil-works/pi-coding-agent'
 import { getEnv } from './env'
-import { getRequestyConfig, updateModelsJson } from './models-json'
+import {
+  diffModels,
+  formatModelsDiffSummary,
+  getRequestyConfig,
+  updateModelsJson,
+  type ModelsDiff,
+} from './models-json'
 import { discoverModels } from './requesty-api'
 import { checkModels, formatHealthSummary, writeHealthCheckLog } from './health-check'
 import { RequestyStatusLoader } from './ui/requesty-status-loader.ts'
@@ -67,10 +73,11 @@ export async function runCommand(args: string, status: StatusReporter, notifier:
 
   try {
     const env = getEnv()
-    const { data, provider } = getRequestyConfig(env)
+    const { data, provider, existingModelIds } = getRequestyConfig(env)
     const models = await discoverModels(provider)
     const modelsMap = new Map(models.map(m => [m.id, m]))
 
+    let diff: ModelsDiff
     let failed = []
     let passing: ProviderModelConfig[]
     let logNote = ''
@@ -89,11 +96,13 @@ export async function runCommand(args: string, status: StatusReporter, notifier:
         const model = modelsMap.get(r.modelId)
         return r.ok && model ? [model] : []
       })
+      diff = diffModels(existingModelIds, passing)
       healthCheckSummary = formatHealthSummary(sortedResults)
-      writeHealthCheckLog(provider, sortedResults, env)
+      writeHealthCheckLog(provider, sortedResults, diff, env)
       logNote = `Full health check log: ${env.health_check_log_path}\n`
     } else {
       passing = models
+      diff = diffModels(existingModelIds, passing)
     }
 
     const shouldUpdate = passing.length > 0 && !dryRun
@@ -102,7 +111,7 @@ export async function runCommand(args: string, status: StatusReporter, notifier:
     }
 
     const writeNote = shouldUpdate ? 'Run /reload to use models.json changes.' : 'models.json was not updated.'
-    const message = `Discovered ${models.length} Requesty model(s).\n${healthCheckSummary}${logNote}${writeNote}`
+    const message = `Discovered ${models.length} Requesty model(s).\n${healthCheckSummary}${formatModelsDiffSummary(diff)}\n${logNote}${writeNote}`
 
     if (failed.length === 0) {
       notifier.notify(message, 'info')

package/src/models-json.ts

@@ -12,6 +12,7 @@ const ProviderSchema = z
     name: z.string().optional(),
     baseUrl: z.string().optional(),
     apiKey: z.string().optional(),
+    models: z.array(z.object({ id: z.string() }).catchall(z.unknown())).optional(),
   })
   .catchall(z.unknown())
 
@@ -33,6 +34,12 @@ type RequestyProvider = ModelsJsonProvider & {
 type RequestyConfig = {
   data: ModelsJson
   provider: RequestyProvider
+  existingModelIds: string[]
+}
+
+export type ModelsDiff = {
+  added: string[]
+  removed: string[]
 }
 
 export function getRequestyConfig(envConfig: Env = getEnv()): RequestyConfig {
@@ -46,6 +53,7 @@ export function getRequestyConfig(envConfig: Env = getEnv()): RequestyConfig {
   const apiKey = envConfig.requesty_api_key
   return {
     data,
+    existingModelIds: (provider.models ?? []).map(m => m.id),
     provider: {
       ...provider,
       name: nonEmptyString(provider.name) ?? DEFAULT_NAME,
@@ -55,6 +63,24 @@ export function getRequestyConfig(envConfig: Env = getEnv()): RequestyConfig {
   }
 }
 
+export function diffModels(previousIds: string[], nextModels: ProviderModelConfig[]): ModelsDiff {
+  const previous = new Set(previousIds)
+  const next = new Set(nextModels.map(m => m.id))
+  return {
+    added: [...next].filter(id => !previous.has(id)).toSorted(),
+    removed: [...previous].filter(id => !next.has(id)).toSorted(),
+  }
+}
+
+export function formatModelsDiffSummary(diff: ModelsDiff): string {
+  return [
+    diff.added.length === 0 ? 'No added models.' : 'Added models:',
+    ...diff.added.map(id => `- ${id}`),
+    diff.removed.length === 0 ? 'No removed models.' : 'Removed models:',
+    ...diff.removed.map(id => `- ${id}`),
+  ].join('\n')
+}
+
 export function updateModelsJson(data: ModelsJson, models: ProviderModelConfig[], envConfig: Env = getEnv()): void {
   const provider = data.providers[envConfig.provider_id]
   data.providers[envConfig.provider_id] = {