Repository navigation
feat(proxy): comprehensive proxy support for all AI providers - #124
Conversation
|
Note Other AI code review bot(s) detectedCodeRabbit has detected other AI code review bot(s) in this pull request and will avoid duplicating their findings in the review comments. This may lead to a less comprehensive review. Important Review skippedAuto incremental reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the You can disable this status message by setting the WalkthroughAdds extensive runtime instrumentation across NeuroLink, Google Vertex, proxy fetch, and provider layers; introduces proxy-based HTTP routing; adds Anthropic-on-Vertex routing and validation; updates provider health checks with Vertex/Anthropic support; adds new NeuroLink in-memory MCP server APIs; adjusts several method signatures (mostly parameter naming/async) and removes redundant generation methods. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant NL as NeuroLink
participant PR as ProviderRegistry
participant MCP as External MCP Server
participant MM as MemoryManager (dyn import)
NL->>NL: Constructor start (C001)
NL->>PR: Configure providers (C002)
PR-->>NL: Providers configured (C004)
NL->>MCP: Init external server (C009)
MCP-->>NL: Setup complete (C010–C012)
NL->>MM: Dynamic import (M004)
MM-->>NL: Import result (M005)
NL-->>NL: Telemetry checkpoints (M007–M015)
sequenceDiagram
autonumber
participant GV as GoogleVertexProvider
participant CFG as createVertexSettings (async)
participant ANT as createVertexAnthropic
participant GML as Google Vertex Models
participant AML as Vertex Anthropic Models
GV->>CFG: Build settings (auth/env) [V003–V007]
CFG-->>GV: Settings (project, region, creds)
GV->>GV: isAnthropicModel?
alt Anthropic
GV->>ANT: Validate & create Anthropic model [V008–V012]
ANT-->>GV: LanguageModelV1 or null
GV->>AML: Stream/generate if model created
else Google
GV->>GML: Create Google Vertex model [V013–V014]
end
sequenceDiagram
autonumber
participant CL as Client
participant PF as createProxyFetch
participant UD as undici.fetch
participant OR as Origin
CL->>PF: fetch(input, init)
PF->>PF: Analyze env/proxy/URL
alt Proxy available
PF->>UD: fetch with ProxyAgent dispatcher
else No proxy
PF->>UD: fetch without dispatcher
end
UD->>OR: HTTP request
OR-->>UD: Response
UD-->>PF: Response
PF-->>CL: Response (or fallback to global fetch on error)
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related PRs
Suggested labels
Poem
✨ Finishing Touches🧪 Generate unit tests
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. CodeRabbit Commands (Invoked using PR/Issue comments)Type Other keywords and placeholders
CodeRabbit Configuration File (
|
✅ Single Commit Policy - COMPLIANTStatus: Policy requirements met • 1 commit • Valid format • Ready for merge 📊 View validation details📝 Commit Details
✅ Validation Results
🤖 Automated validation by NeuroLink Single Commit Enforcement |
🤖 AI Review & Build Compliance ✅Status: AI analysis complete • Build rules validated • Ready for review 📊 View detailed analysis results🛡️ Analysis Complete
📋 Ready for Merge When
🤖 AI analysis complete - check individual code comments for specific feedback |
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
There was a problem hiding this comment.
Pull Request Overview
This PR introduces comprehensive proxy support across all AI providers, enhanced authentication mechanisms for Google Cloud services, and extensive logging infrastructure for debugging corporate firewall environments.
Key Changes:
- Universal Proxy Integration: All AI providers now support proxy connections via
createProxyFetch()function - Enhanced Google Vertex Authentication: Dual authentication methods with file existence validation and runtime credentials file creation
- Comprehensive Logging: Exhaustive debug logging for proxy configurations, network requests, and authentication flows
Reviewed Changes
Copilot reviewed 9 out of 9 changed files in this pull request and generated 12 comments.
Show a summary per file
| File | Description |
|---|---|
| tools/automation/buildSystem.js | Code formatting improvements with multi-line function calls |
| src/lib/utils/providerHealth.ts | Enhanced Vertex AI authentication validation and comprehensive Anthropic support checking |
| src/lib/proxy/proxyFetch.ts | Detailed proxy request logging with comprehensive error analysis |
| src/lib/providers/openaiCompatible.ts | Proxy support integration and code cleanup |
| src/lib/providers/litellm.ts | Proxy support integration and unused import removal |
| src/lib/providers/googleVertex.ts | Major authentication overhaul with dual provider architecture and extensive logging |
| src/lib/providers/amazonBedrock.ts | Proxy support integration |
| src/lib/neurolink.ts | Comprehensive initialization and stream processing logging |
| src/lib/mcp/toolDiscoveryService.ts | Minor formatting improvements |
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 10
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (5)
src/lib/mcp/toolDiscoveryService.ts (1)
751-757: Avoid logging entire tool outputs; redact and truncate to prevent sensitive data leakage and log bloat
fullResponse: resultcan dump PII, API tokens, or large payloads to logs. Redact common secret-looking keys and cap size.- mcpLogger.debug("[ToolDiscoveryService] Tool response received", { - type: typeof result, - isArray: Array.isArray(result), - isObject: isObject(result), - hasKeys: isObject(result) ? Object.keys(result as object).length : 0, - fullResponse: result, // Log the complete response, not a truncated sample - }); + // Redact and limit payload to avoid leaking sensitive data or flooding logs + const safePreview = (() => { + try { + const MAX = 2048; + const json = + typeof result === "string" + ? result + : JSON.stringify(result, (_k, v) => { + // redact common secret-like fields + if ( + typeof _k === "string" && + /token|secret|authorization|api[_-]?key|password/i.test(_k) + ) { + return "[REDACTED]"; + } + return v; + }); + return json.length > MAX ? json.slice(0, MAX) + "…[truncated]" : json; + } catch { + return String(result).slice(0, 512); + } + })(); + mcpLogger.debug("[ToolDiscoveryService] Tool response received", { + type: typeof result, + isArray: Array.isArray(result), + isObject: isObject(result), + hasKeys: isObject(result) ? Object.keys(result as object).length : 0, + preview: safePreview, + });Optionally, extract the redaction logic into a shared utility if used elsewhere.
src/lib/neurolink.ts (1)
1391-1527: MCP retry loop double-increments attempts; reduces total retries and misreports attempt numbers
mcpAttemptsis incremented in the catch and again after the try/catch, causing off-by-one behavior and fewer actual attempts than configured. It also logs incorrect attempt indices. Refactor to a for-loop or ensure a single increment per iteration.- while (mcpAttempts <= maxMcpRetries) { - try { + const maxAttempts = maxMcpRetries + 1; + for (let attempt = 1; attempt <= maxAttempts; attempt++) { + try { // 🚀 EXHAUSTIVE LOGGING POINT G007: MCP GENERATION ATTEMPT const mcpAttemptStartTime = process.hrtime.bigint(); logger.debug(`[NeuroLink] 🎯 LOG_POINT_G007_MCP_ATTEMPT_START`, { logPoint: "G007_MCP_ATTEMPT_START", generateInternalId, timestamp: new Date().toISOString(), elapsedMs: Date.now() - generateInternalStartTime, elapsedNs: ( process.hrtime.bigint() - generateInternalHrTimeStart ).toString(), mcpAttemptStartTimeNs: mcpAttemptStartTime.toString(), - currentAttempt: mcpAttempts + 1, - maxAttempts: maxMcpRetries + 1, - isFirstAttempt: mcpAttempts === 0, - isLastAttempt: mcpAttempts === maxMcpRetries, - attemptType: mcpAttempts === 0 ? "INITIAL" : "RETRY", - message: `Attempting MCP generation (attempt ${mcpAttempts + 1}/${maxMcpRetries + 1})`, + currentAttempt: attempt, + maxAttempts, + isFirstAttempt: attempt === 1, + isLastAttempt: attempt === maxAttempts, + attemptType: attempt === 1 ? "INITIAL" : "RETRY", + message: `Attempting MCP generation (attempt ${attempt}/${maxAttempts})`, }); logger.debug( - `[${functionTag}] Attempting MCP generation (attempt ${mcpAttempts + 1}/${maxMcpRetries + 1})...`, + `[${functionTag}] Attempting MCP generation (attempt ${attempt}/${maxAttempts})...`, ); const mcpResult = await this.tryMCPGeneration(options); // … unchanged … if (mcpResult && mcpResult.content) { // success path … return mcpResult; } else { // analyze empty content path … if ( mcpResult && mcpResult.toolExecutions && mcpResult.toolExecutions.length > 0 ) { // success-with-tools path … return mcpResult; } } - } catch (error) { - mcpAttempts++; + } catch (error) { logger.debug( - `[${functionTag}] MCP generation failed on attempt ${mcpAttempts}/${maxMcpRetries + 1}`, + `[${functionTag}] MCP generation failed on attempt ${attempt}/${maxAttempts}`, { error: error instanceof Error ? error.message : String(error), - willRetry: mcpAttempts <= maxMcpRetries, + willRetry: attempt < maxAttempts, }, ); - // If this was the last attempt, break and fall back - if (mcpAttempts > maxMcpRetries) { + // If this was the last attempt, break and fall back + if (attempt >= maxAttempts) { logger.debug( `[${functionTag}] All MCP attempts exhausted, falling back to direct generation`, ); break; } // Small delay before retry to allow transient issues to resolve await new Promise((resolve) => setTimeout(resolve, 500)); - } - - mcpAttempts++; - } + } + }src/lib/utils/providerHealth.ts (1)
1398-1412: Include Azure in health checksThe default preferredProviders in getBestHealthyProvider includes
"azure", but checkAllProvidersHealth only iterates over:
- AIProviderName.ANTHROPIC
- AIProviderName.OPENAI
- AIProviderName.VERTEX
- AIProviderName.GOOGLE_AI
- AIProviderName.BEDROCK
- AIProviderName.OLLAMA
As a result, Azure will never be health-checked and may be returned without verification. Since Azure is a first-class, registered provider (see providerRegistry and AzureOpenAIProvider implementation), we should add it to the health-check list.
• File: src/lib/utils/providerHealth.ts (around line 1451)
• Modify the providers array in checkAllProvidersHealth:const providers: AIProviderName[] = [ AIProviderName.ANTHROPIC, AIProviderName.OPENAI, AIProviderName.VERTEX, AIProviderName.GOOGLE_AI, AIProviderName.BEDROCK, + AIProviderName.AZURE, AIProviderName.OLLAMA, ];If Azure is intentionally excluded from quick health checks, instead remove
"azure"from the default list in getBestHealthyProvider. But given Azure’s full implementation elsewhere, adding it here ensures consistency.src/lib/proxy/proxyFetch.ts (1)
149-155: Preserve Request semantics; don’t coerce input to stringCoercing input to string breaks Request objects (method/body/headers). Pass input through unchanged.
Apply this diff:
- const response = await undici.fetch( - input as string, - { + const response = await undici.fetch( + input as any, + { ...init, dispatcher: dispatcher, } as unknown as import("undici").RequestInit, );src/lib/providers/amazonBedrock.ts (1)
24-27: Remove account-specific ARN default; it will break for all other accountsThe fallback to arn:aws:bedrock:...:225681119357:... hard-codes a specific account’s inference profile and will fail universally.
Use an env-only configuration or a broadly available model identifier. Example fix:
@@ - return ( - process.env.BEDROCK_MODEL || - process.env.BEDROCK_MODEL_ID || - "arn:aws:bedrock:us-east-2:225681119357:inference-profile/us.anthropic.claude-3-7-sonnet-20250219-v1:0" - ); + const model = + process.env.BEDROCK_MODEL || process.env.BEDROCK_MODEL_ID; + if (!model) { + throw new Error( + "BEDROCK_MODEL (or BEDROCK_MODEL_ID) is required. Example: 'anthropic.claude-3-haiku-20240307-v1:0' or a valid inference profile ARN.", + ); + } + return model;
🧹 Nitpick comments (14)
src/lib/mcp/toolDiscoveryService.ts (1)
741-748: Comment and behavior mismatch: we never throw on null/undefined resultsThe comment says "Only throw for truly invalid responses (null/undefined)" but the code logs and returns without throwing. Either fix the comment or introduce a throw. Given the intended permissive behavior, I'd update the comment to avoid confusion.
- // Only throw for truly invalid responses (null/undefined) + // For null/undefined responses, do not throw; treat as empty and let the AI handle it if (isNullish(result)) { mcpLogger.debug( "[ToolDiscoveryService] Tool returned null/undefined, treating as empty response", ); // Even null responses can be valid for some tools - don't throw return; }tools/automation/buildSystem.js (1)
380-382: Nit: prefer a function to echo the skip-build message for consistency/testingCurrent constant is fine; if you later add color or localization, a helper would centralize it. No change required now.
src/lib/neurolink.ts (1)
631-649: Optional: make initializeMCP concurrency-safeIf generate() and stream() call initializeMCP concurrently, both paths may run heavy initialization. Guard with a shared in-flight promise.
Sketch:
// class field private mcpInitInFlight?: Promise<void>; // in initializeMCP() entry if (this.mcpInitialized) return; if (this.mcpInitInFlight) { await this.mcpInitInFlight; return; } this.mcpInitInFlight = (async () => { try { // existing initialization … this.mcpInitialized = true; } finally { this.mcpInitInFlight = undefined; } })(); await this.mcpInitInFlight;src/lib/providers/googleVertex.ts (3)
156-175: Auth fallback flow is solid, but consider early-returning after principal-file success and reducing duplicate env parsingYou already return early if the ADC file exists. For readability and to avoid parsing the env twice, consider structuring as: (1) ADC file exists → return; (2) individual env creds → return settings with googleAuthOptions; (3) final warn.
Also applies to: 176-237
585-656: Create Vertex instance/model with proxy-aware fetch is correct; add small hardening for network diagnosticsCurrent diagnostics include DNS servers and interfaces; that’s very verbose. Keep at debug level (as you do) and consider gating behind an env flag to reduce noise in prod. No code change required now.
Also applies to: 657-788
59-64: Model default string may drift; consider centralizing latest default in configHardcoding "gemini-2.5-flash" is fine today. If you already have a provider config manager (you do), consider pulling the default from there to avoid future churn.
src/lib/utils/providerHealth.ts (2)
486-496: Clarify Vertex API key env var lookup to avoid misleading consumersgetApiKeyEnvironmentVariable returns "GOOGLE_APPLICATION_CREDENTIALS" for Vertex, but Vertex auth supports both a credentials file and GOOGLE_AUTH_CLIENT_EMAIL/GOOGLE_AUTH_PRIVATE_KEY. Returning a single var can mislead downstream code/tools.
Two options:
- Return an empty string for Vertex and rely on the Vertex-specific branch in checkApiKeyValidity.
- Or return a descriptive pseudo-key like "VERTEX_AUTH" and update any messaging to explain both auth paths.
1238-1248: Hardcoded regional support list will drift; consider single source of truthThe supportedRegions array will stale over time. Pulling this list from a config module or an env override avoids frequent code changes and mismatches.
Expose supported regions in a constants file or accept a comma-separated env var (VERTEX_ANTHROPIC_SUPPORTED_REGIONS) and default to the current list.
src/lib/providers/openaiCompatible.ts (2)
90-91: Avoid re-instantiating ProxyAgent per call; reuse a single proxy-aware fetchcreateProxyFetch() builds a ProxyAgent; creating a new one per request adds overhead. Reuse a single instance across the provider.
Apply this diff:
@@ - this.customOpenAI = createOpenAI({ + this.customOpenAI = createOpenAI({ baseURL: this.config.baseURL, apiKey: this.config.apiKey, - fetch: createProxyFetch(), + fetch: this.proxyFetch, }); @@ - const proxyFetch = createProxyFetch(); - const response = await proxyFetch(modelsUrl, { + const response = await this.proxyFetch(modelsUrl, { headers: { Authorization: `Bearer ${this.config.apiKey}`, "Content-Type": "application/json", }, });And add this class field near other fields:
private readonly proxyFetch: typeof fetch = createProxyFetch();Also applies to: 300-302
141-145: Auto-discovery picks the first model; consider preferring stable/default modelsSelecting the first returned model can be non-deterministic across providers. If the list includes embeddings or legacy models first, users get surprising results.
Prefer a small allowlist (e.g., gpt-4o, gpt-4o-mini, gpt-3.5-turbo) if present; else fall back to current behavior.
src/lib/providers/litellm.ts (2)
67-68: Reuse a single proxy-aware fetch instead of creating one per requestSame rationale as openaiCompatible: reduce overhead by reusing one ProxyAgent.
Apply this diff:
@@ - const customOpenAI = createOpenAI({ + const customOpenAI = createOpenAI({ baseURL: config.baseURL, apiKey: config.apiKey, - fetch: createProxyFetch(), + fetch: this.proxyFetch, }); @@ - const proxyFetch = createProxyFetch(); - const response = await proxyFetch(modelsUrl, { + const response = await this.proxyFetch(modelsUrl, { method: "GET", headers: { Authorization: `Bearer ${config.apiKey}`, "Content-Type": "application/json", }, signal: controller.signal, });Add this class field alongside
private model: LanguageModelV1;:private readonly proxyFetch: typeof fetch = createProxyFetch();Also applies to: 303-305
271-280: Make fallback models configurable for easier ops overridesYou already read LITELLM_FALLBACK_MODELS; consider logging the env source and trimming/validating entries to reduce surprise at runtime.
No code change required; optional: log which source (env vs default) is used and ignore empty strings after split/trim.
src/lib/proxy/proxyFetch.ts (2)
54-57: Honor NO_PROXY and ALL_PROXY for correctness in mixed-network environmentsCurrent selection ignores NO_PROXY and ALL_PROXY. Respecting them avoids proxying localhost/cluster endpoints and supports a single ALL_PROXY setting.
Sketch:
const allProxy = process.env.ALL_PROXY || process.env.all_proxy; const noProxy = (process.env.NO_PROXY || process.env.no_proxy || "") .split(",") .map((s) => s.trim()) .filter(Boolean); const shouldBypass = noProxy.some((pattern) => url.hostname.endsWith(pattern)); const selected = shouldBypass ? undefined : (url.protocol === "https:" ? httpsProxy : httpProxy) || allProxy;Then use selected for ProxyAgent choice.
114-175: Cache ProxyAgent per proxy URL to reduce overheadCreating a ProxyAgent for every request is unnecessary. Memoize per proxyUrl.
Minimal change:
const agentCache = new Map<string, import("undici").ProxyAgent>(); // ... let dispatcher = agentCache.get(proxyUrl); if (!dispatcher) { dispatcher = new ProxyAgent(proxyUrl); agentCache.set(proxyUrl, dispatcher); }
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (9)
src/lib/mcp/toolDiscoveryService.ts(2 hunks)src/lib/neurolink.ts(9 hunks)src/lib/providers/amazonBedrock.ts(4 hunks)src/lib/providers/googleVertex.ts(8 hunks)src/lib/providers/litellm.ts(5 hunks)src/lib/providers/openaiCompatible.ts(5 hunks)src/lib/proxy/proxyFetch.ts(4 hunks)src/lib/utils/providerHealth.ts(9 hunks)tools/automation/buildSystem.js(3 hunks)
🧰 Additional context used
🧬 Code graph analysis (8)
src/lib/mcp/toolDiscoveryService.ts (1)
src/lib/utils/typeUtils.ts (1)
isNullish(97-99)
src/lib/proxy/proxyFetch.ts (2)
src/lib/utils/logger.ts (2)
logger(198-237)error(134-136)tools/testing/proxy_server.js (1)
URL(29-29)
src/lib/providers/litellm.ts (1)
src/lib/proxy/proxyFetch.ts (1)
createProxyFetch(12-209)
src/lib/providers/openaiCompatible.ts (2)
src/lib/proxy/proxyFetch.ts (1)
createProxyFetch(12-209)src/lib/types/typeAliases.ts (1)
ZodUnknownSchema(18-18)
src/lib/providers/amazonBedrock.ts (4)
src/lib/proxy/proxyFetch.ts (1)
createProxyFetch(12-209)src/lib/types/typeAliases.ts (1)
ZodUnknownSchema(18-18)src/lib/types/streamTypes.ts (1)
StreamResult(144-178)src/lib/utils/messageBuilder.ts (1)
buildMessagesArray(16-64)
src/lib/neurolink.ts (9)
src/lib/utils/logger.ts (3)
logger(198-237)error(134-136)mcpLogger(240-240)src/lib/factories/providerRegistry.ts (1)
ProviderRegistry(26-276)src/lib/utils/conversationMemoryUtils.ts (1)
applyConversationMemoryDefaults(22-33)src/lib/core/conversationMemoryManager.ts (1)
ConversationMemoryManager(20-214)src/lib/mcp/externalServerManager.ts (1)
ExternalServerManager(155-1468)src/lib/utils/performance.ts (1)
MemoryManager(109-185)src/lib/mcp/toolRegistry.ts (2)
toolRegistry(827-827)ToolInfo(831-831)src/lib/mcp/servers/agent/directToolsServer.ts (1)
directToolsServer(17-23)src/lib/mcp/contracts/mcpContract.ts (1)
ToolInfo(50-58)
src/lib/utils/providerHealth.ts (2)
src/lib/index.ts (1)
AIProviderName(15-15)src/lib/utils/logger.ts (2)
logger(198-237)error(134-136)
src/lib/providers/googleVertex.ts (4)
src/lib/proxy/proxyFetch.ts (1)
createProxyFetch(12-209)src/lib/utils/logger.ts (2)
logger(198-237)error(134-136)src/lib/utils/timeout.ts (1)
createTimeoutController(436-469)src/lib/utils/messageBuilder.ts (1)
buildMessagesArray(16-64)
🔇 Additional comments (5)
tools/automation/buildSystem.js (2)
350-353: LGTM: clearer deploy command construction with explicit default for forceRebuildMoving the boolean default and formatting the call improves readability without changing behavior.
468-470: LGTM: CLI usage text wrapping improves readabilityUser help is clearer; no behavior changes.
src/lib/providers/googleVertex.ts (2)
365-410: Dual-path model routing and instrumentation look goodAnthropic-on-Vertex path selection and detailed logging are helpful for ops. The fallback to Google model when Anthropic is unavailable is sensible.
Also applies to: 432-554
869-982: Stream timeout handling is consistent; fix missing AbortSignal in AmazonBedrockProviderThe other providers correctly pass
abortSignal(viatimeoutController.controller.signal) down intostreamTextor HTTP fetch calls, but inAmazonBedrockProvider.executeStreamwe neither set up a timeout controller nor pass anabortSignaltostreamText. As a result, Bedrock streams can’t be aborted by the user or timeout.Locations to update:
- src/lib/providers/amazonBedrock.ts
Suggested diff:
--- a/src/lib/providers/amazonBedrock.ts +++ b/src/lib/providers/amazonBedrock.ts @@ protected async executeStream( - const result = streamText({ - model: this.model, - messages: messages, - maxTokens: options.maxTokens || DEFAULT_MAX_TOKENS, - temperature: options.temperature, - }); + // 🚀 Add timeout controller & propagate abortSignal + const timeout = this.getTimeout(options); + const timeoutController = createTimeoutController( + timeout, + this.providerName, + "stream", + ); + const result = streamText({ + model: this.model, + messages: messages, + maxTokens: options.maxTokens || DEFAULT_MAX_TOKENS, + temperature: options.temperature, + abortSignal: timeoutController.controller.signal, + }); + timeoutController.cleanup();Please apply this change to ensure the Bedrock stream can be properly aborted.
Likely an incorrect or invalid review comment.
src/lib/providers/amazonBedrock.ts (1)
71-76: Good: Bedrock client wired to proxy-aware fetchWiring fetch: createProxyFetch() into the AWS client aligns with the PR objective and will help customers behind corporate proxies.
c745e7d to
74326a1
Compare
🤖 AI Review & Build Compliance ✅Status: AI analysis complete • Build rules validated • Ready for review 📊 View detailed analysis results🛡️ Analysis Complete
📋 Ready for Merge When
🤖 AI analysis complete - check individual code comments for specific feedback |
74326a1 to
49c5ed7
Compare
🤖 AI Review & Build Compliance ✅Status: AI analysis complete • Build rules validated • Ready for review 📊 View detailed analysis results🛡️ Analysis Complete
📋 Ready for Merge When
🤖 AI analysis complete - check individual code comments for specific feedback |
- **Enhanced Proxy Infrastructure**: Comprehensive proxy configuration system with exhaustive environment variable detection and logging - **Universal Provider Integration**: All AI providers (Vertex, Bedrock, OpenAI-Compatible, LiteLLM) now support proxy connections via createProxyFetch() - **Robust Authentication**: Improved Google Cloud authentication with dual methods (service account files + individual env vars) and proper file existence validation - **Enhanced Health Monitoring**: Provider health checks now include proxy-aware connectivity testing and comprehensive validation - **ESM Compatibility**: Resolved CommonJS/ESM compatibility issues for Node.js module imports - **Extensive Logging**: Detailed proxy configuration and request tracking for debugging corporate firewall environments
49c5ed7 to
9a4e8e1
Compare
🤖 AI Review & Build Compliance ✅Status: AI analysis complete • Build rules validated • Ready for review 📊 View detailed analysis results🛡️ Analysis Complete
📋 Ready for Merge When
🤖 AI analysis complete - check individual code comments for specific feedback |
Pull Request
Description
Type of Change
Related Issues
Changes Made
AI Provider Impact
Component Impact
Testing
Test Environment
Performance Impact
Breaking Changes
Screenshots/Demo
Checklist
Additional Notes
Summary by CodeRabbit
New Features
Improvements
Refactor
Style