fix: config consistency refactor

__tests__/anthropic-senders/line-comments-sender.test.ts

@@ -1,5 +1,6 @@
 import { beforeEach, describe, expect, it, vi } from 'vitest'
 import { anthropicLineCommentsSender } from '../../src/senders/providers/anthropic/line-comments-sender.ts'
+import { _resetRuntimeConfigCacheForTests } from '../../src/core/utils/runtime-config.ts'
 
 // Mock the Anthropic SDK
 vi.mock('@anthropic-ai/sdk', () => {
@@ -127,6 +128,7 @@ describe('anthropicLineCommentsSender', () => {
 
   beforeEach(async () => {
     vi.clearAllMocks()
+    _resetRuntimeConfigCacheForTests()
 
     // Get the mocked Anthropic constructor
     const AnthropicConstructor = (await import('@anthropic-ai/sdk'))

__tests__/comment-handlers/discussion-handler.test.ts

@@ -114,6 +114,8 @@ describe('handleDiscussionReply', () => {
     expect(client.replyToReviewComment).toHaveBeenCalledTimes(1)
     const [[, , body]] = (client.replyToReviewComment as Mock).mock.calls
     expect(typeof body).toBe('string')
-    expect(body).toContain('I could not generate a confident, useful automated reply')
+    expect(body).toContain(
+      'I could not generate a confident, useful automated reply'
+    )
   })
 })

__tests__/config-loader.test.ts

@@ -45,12 +45,14 @@ describe('getAppConfig / LLM provider resolution', () => {
     expect(cfg.llmProvider).toBe('anthropic')
   })
 
-  it('uses LLM_PROVIDER when config.json is missing', async () => {
+  it('ignores LLM_PROVIDER when config.json is missing (provider resolved at runtime layer)', async () => {
     process.env.LLM_PROVIDER = 'openai'
     vi.mocked(fs.readFile).mockRejectedValueOnce(new Error('no file'))
 
     const cfg = await getAppConfig()
-    expect(cfg.llmProvider).toBe('openai')
+    // getAppConfig now only reflects config.json + defaults; provider env
+    // resolution is handled in the runtime-config layer.
+    expect(cfg.llmProvider).toBe('anthropic')
   })
 
   it('prefers config.json llmProvider over env', async () => {
@@ -65,7 +67,7 @@ describe('getAppConfig / LLM provider resolution', () => {
     expect(cfg.llmProvider).toBe('anthropic')
   })
 
-  it('falls back to env when config.json omits llmProvider', async () => {
+  it('does not fall back to env when config.json omits llmProvider (runtime layer handles env)', async () => {
     process.env.LLM_PROVIDER = 'openai'
 
     const fileCfg: Partial<RevuAppConfig> = {
@@ -74,7 +76,9 @@ describe('getAppConfig / LLM provider resolution', () => {
     vi.mocked(fs.readFile).mockResolvedValueOnce(JSON.stringify(fileCfg))
 
     const cfg = await getAppConfig()
-    expect(cfg.llmProvider).toBe('openai')
+    // getAppConfig remains config.json + defaults; env-based provider
+    // fallback is handled by runtime-config.
+    expect(cfg.llmProvider).toBe('anthropic')
   })
 
   it('logs a warning and ignores invalid LLM_PROVIDER', async () => {

__tests__/openai-senders/line-comments-sender.test.ts

@@ -1,5 +1,6 @@
 import { describe, it, expect, beforeEach, vi, afterEach } from 'vitest'
 import { openaiLineCommentsSender } from '../../src/senders/providers/openai/line-comments-sender.ts'
+import { _resetRuntimeConfigCacheForTests } from '../../src/core/utils/runtime-config.ts'
 
 // Mock the OpenAI SDK
 const createMock = vi.fn()
@@ -17,6 +18,7 @@ describe('openaiLineCommentsSender', () => {
   beforeEach(async () => {
     vi.resetModules()
     createMock.mockReset()
+    _resetRuntimeConfigCacheForTests()
     // Ensure env is set for tests
     process.env.OPENAI_API_KEY = 'test-openai-key'
     delete process.env.OPENAI_MODEL

src/comment-handlers/discussion-handler.ts

@@ -8,6 +8,7 @@ import {
   simpleHash
 } from '../utils/compute-cache.ts'
 import { logSystemError, logSystemWarning } from '../utils/logger.ts'
+import { getRuntimeConfig } from '../core/utils/runtime-config.ts'
 
 export interface ThreadMessage {
   author: string
@@ -113,11 +114,14 @@ export async function handleDiscussionReply(params: DiscussionHandlerParams) {
       ? userReplyBody.slice(0, MAX_REPLY_HASH_CHARS)
       : userReplyBody
   const bodyHash = simpleHash(truncatedForHash, 16)
-  const provider = process.env.LLM_PROVIDER || 'anthropic'
+
+  const runtime = await getRuntimeConfig()
+  const provider = runtime.llm.provider
   const modelForKey =
     provider === 'openai'
-      ? process.env.OPENAI_MODEL || 'gpt-5'
-      : process.env.ANTHROPIC_MODEL || 'claude-sonnet-4-5-20250929'
+      ? runtime.llm.openai.model
+      : runtime.llm.anthropic.model
+
   const cacheKey = buildDiscussionCacheKey({
     owner,
     repo,
@@ -194,7 +198,7 @@ export async function handleDiscussionReply(params: DiscussionHandlerParams) {
 
   let acquired = false
   if (hasLockSupport) {
-    const lockTtl = Number(process.env.DISCUSSION_LOCK_TTL_SECONDS || 240)
+    const lockTtl = runtime.discussion.lockTtlSeconds
     acquired = await cacheAny.tryAcquireLock(lockKey, lockTtl)
   }
 
@@ -282,7 +286,8 @@ export async function handleDiscussionReply(params: DiscussionHandlerParams) {
     userReplyBody: replyForPrompt,
     history,
     relevantFilePath,
-    diffHunk
+    diffHunk,
+    maxCharsPerFile: runtime.discussion.maxFileContentChars
   })
 
   // Generate assistant reply (concise)

src/core/utils/config-loader.ts

@@ -2,7 +2,6 @@ import * as fs from 'fs/promises'
 import * as path from 'path'
 import { DEFAULT_APP_CONFIG, type RevuAppConfig } from '../../types/config.ts'
 import { logSystemWarning } from '../../utils/logger.ts'
-import { applyRevuAppEnvOverrides } from './env-config.ts'
 /**
  * Reads the application configuration from config.json
  */
@@ -23,50 +22,15 @@ export async function getAppConfig(): Promise<RevuAppConfig> {
       ...fileConfig
     }
 
-    applyRevuAppEnvOverrides(merged, fileConfig, [
-      {
-        key: 'llmProvider',
-        envVar: 'LLM_PROVIDER',
-        parse: (raw) => raw.toLowerCase() as RevuAppConfig['llmProvider'],
-        validate: (value) => value === 'anthropic' || value === 'openai',
-        onInvalid: (raw) => {
-          logSystemWarning(
-            new Error(
-              'Invalid LLM_PROVIDER env var, expected "anthropic" or "openai"'
-            ),
-            { context_msg: `value="${raw}"` }
-          )
-        }
-      }
-    ])
-
     cachedConfig = merged
     return cachedConfig
   } catch (error) {
     logSystemWarning('Failed to read config.json, using defaults:', error)
 
-    const fileConfig: Partial<RevuAppConfig> = {}
     const merged: RevuAppConfig = {
       ...DEFAULT_APP_CONFIG
     }
 
-    applyRevuAppEnvOverrides(merged, fileConfig, [
-      {
-        key: 'llmProvider',
-        envVar: 'LLM_PROVIDER',
-        parse: (raw) => raw.toLowerCase() as RevuAppConfig['llmProvider'],
-        validate: (value) => value === 'anthropic' || value === 'openai',
-        onInvalid: (raw) => {
-          logSystemWarning(
-            new Error(
-              'Invalid LLM_PROVIDER env var, expected "anthropic" or "openai"'
-            ),
-            { context_msg: `value="${raw}"` }
-          )
-        }
-      }
-    ])
-
     cachedConfig = merged
     return cachedConfig
   }