Fix backend startup errors and API proxy issues for external IP access#565
Fix backend startup errors and API proxy issues for external IP access#565croakingtoad wants to merge 1 commit intocoleam00:mainfrom
Conversation
This PR fixes critical issues preventing Archon from working correctly when accessed via external IP addresses (not localhost). ## Issues Fixed 1. **Backend Service Startup Failure** - Frontend showed persistent error popup 2. **Configuration check failures** - "Failed to fetch" errors in Settings 3. **Projects schema verification** - Unable to verify projects schema warning ## Root Causes 1. **Missing `ready` field in health endpoint**: The `/api/health` endpoint in knowledge_api.py was missing the `ready: true` field that the frontend expected 2. **Direct backend port access instead of proxy**: Frontend services were constructing URLs like `http://EXTERNAL_IP:8181` instead of using relative URLs through Vite proxy 3. **Hardcoded API URLs at service instantiation**: Service classes were setting `baseUrl` once at startup, not dynamically ## Changes Made ### Backend (Python) - `python/src/server/api_routes/knowledge_api.py`: Added `ready`, `credentials_loaded`, and `schema_valid` fields to health response ### Frontend (React/TypeScript) - `archon-ui-main/src/config/api.ts`: Changed `getApiUrl()` to always return empty string in development (forces proxy usage) - `archon-ui-main/src/components/layouts/MainLayout.tsx`: Use relative URL `/api/health` instead of constructing full URL - `archon-ui-main/src/components/settings/FeaturesSection.tsx`: Use relative URL for projects health check - `archon-ui-main/src/services/credentialsService.ts`: Replace all API calls with relative URLs - `archon-ui-main/src/services/mcpClientService.ts`: Replace all API calls with relative URLs - `archon-ui-main/src/services/bugReportService.ts`: Use relative URL for bug report endpoint - `archon-ui-main/src/services/testService.ts`: Set API_BASE_URL to empty string ## Testing - Verified backend health endpoint returns correct format - Confirmed all API calls go through Vite proxy (port 3737) - Tested Settings page loads without errors - Verified Projects toggle works correctly ## Impact This fix ensures Archon works correctly when: - Deployed on cloud servers (DigitalOcean, AWS, etc.) - Accessed via external IP addresses - Running behind reverse proxies - Used in development with non-localhost access 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
WalkthroughFrontend fetches switch to relative /api URLs routed via the Vite proxy; API config adds helpers and public constants for API/WS resolution; MCP client service adopts uniform error parsing and keeps Archon integration; backend knowledge health response includes readiness fields. No public signatures changed. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant UI as UI (Browser)
participant Vite as Vite Dev Proxy
participant API as Backend API
UI->>Vite: GET /api/health
Vite->>API: Forward /api/health
API-->>Vite: 200 OK {status, ready, credentials_loaded, schema_valid}
Vite-->>UI: 200 OK ...
note over UI: Health checks now use relative /api paths via proxy
sequenceDiagram
autonumber
participant UI as UI Components
participant MCP as mcpClientService
participant API as /api/mcp/...
UI->>MCP: callClientTool(params)
MCP->>API: POST /api/mcp/clients/:id/tools/:tool/call
alt success
API-->>MCP: 200 OK {result}
MCP-->>UI: result
else non-ok
API-->>MCP: 4xx/5xx {detail:"message"}
MCP->>MCP: parse JSON, throw Error(detail || default)
MCP-->>UI: throw Error(...)
note over MCP,UI: Unified error parsing across MCP endpoints
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Poem
✨ Finishing Touches
🧪 Generate unit tests
🪧 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 (
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
archon-ui-main/src/services/mcpClientService.ts (2)
412-416: Robust host resolution: new URL may throw for relative API URLs.If getApiUrl() returns a relative path (e.g., '/api'),
new URL(apiUrl)throws. Resolve the base from window.location when the API URL isn’t absolute.- const apiUrl = getApiUrl(); - const url = new URL(apiUrl || `http://${window.location.hostname}:${mcpPort}`); - const mcpUrl = `${url.protocol}//${url.hostname}:${mcpPort}/mcp`; + const apiUrl = getApiUrl(); + const isAbsolute = !!apiUrl && /^https?:\/\//i.test(apiUrl); + const base = isAbsolute ? new URL(apiUrl as string) : new URL(window.location.origin); + const mcpUrl = `${base.protocol}//${base.hostname}:${mcpPort}/mcp`;
375-377: Ensure return type; don’t leak raw rejection reasons.Return a typed fallback object when a status promise is rejected.
- return statuses.map((result) => - result.status === 'fulfilled' ? result.value : result.reason - ); + return statuses.map((result, index) => + result.status === 'fulfilled' + ? result.value + : { + client: clients[index], + status: { + client_id: clients[index].id, + status: 'error', + last_seen: null, + last_error: (result as PromiseRejectedResult).reason?.message || 'Unknown error', + is_active: false + } + } + );
🧹 Nitpick comments (17)
archon-ui-main/src/services/bugReportService.ts (2)
164-170: Good: route bug reports via proxy-relative endpoint. Add a short timeout.The switch to '/api/bug-report/github' fixes CORS/proxy issues. Add a timeout to avoid hanging UX.
- const response = await fetch('/api/bug-report/github', { + const response = await fetch('/api/bug-report/github', { method: 'POST', headers: { 'Content-Type': 'application/json', }, body: JSON.stringify(requestData), + signal: AbortSignal.timeout(8000), });
7-7: Remove unused import.
getApiUrlis no longer used.-import { getApiUrl } from '../config/api'; +// (removed) import { getApiUrl } from '../config/api';archon-ui-main/src/components/layouts/MainLayout.tsx (1)
48-49: Good: force health check through Vite proxy. Consider centralizing base path.Using '/api/health' is correct. Prefer
getApiBasePath()to avoid scattering the '/api' literal and ease future rewrites.+import { getApiBasePath } from '../../config/api'; ... - const response = await fetch('/api/health', { + const response = await fetch(`${getApiBasePath()}/health`, { method: 'GET', signal: controller.signal });archon-ui-main/src/services/testService.ts (2)
45-47: Avoid local API base duplication; use centralized config.Defining
API_BASE_URL = ''here diverges fromconfig/api.tsand risks drift. UsegetApiBasePath()and pass endpoints without the '/api' prefix.-import { getApiUrl, getWebSocketUrl } from '../config/api'; +import { getApiBasePath, getWebSocketUrl } from '../config/api'; ... -// Always use relative URLs to go through Vite proxy -const API_BASE_URL = ''; +const API_BASE_URL = getApiBasePath();Follow-up (illustrative) to drop '/api' prefix at call sites:
-const response = await callAPI<TestExecution>('/api/tests/mcp/run', { +const response = await callAPI<TestExecution>('/tests/mcp/run', {If you prefer to keep endpoints with '/api' prefixed, then set
const API_BASE_URL = ''via a single export in config and import it here to avoid duplication. I can prep a small PR to standardize either approach.
42-42: Remove unused import.
getApiUrlisn’t used in this module.-import { getApiUrl, getWebSocketUrl } from '../config/api'; +import { getWebSocketUrl } from '../config/api';archon-ui-main/src/services/credentialsService.ts (5)
81-87: Unify error handling and surface HTTP details.Route network/Fetch failures through handleCredentialError and include status/text for non-OK responses to keep UX consistent with update/create/delete.
Apply:
async getAllCredentials(): Promise<Credential[]> { - const response = await fetch('/api/credentials'); - if (!response.ok) { - throw new Error("Failed to fetch credentials"); - } - return response.json(); + try { + const response = await fetch('/api/credentials'); + if (!response.ok) { + const errorText = await response.text().catch(() => ''); + throw new Error(`HTTP ${response.status}: ${errorText || response.statusText}`); + } + return response.json(); + } catch (error) { + throw this.handleCredentialError(error, 'Fetching credentials'); + } }
90-93: Encode path segment and standardize error handling.If category ever contains special chars, the path will break. Also mirror the improved error handling.
- const response = await fetch( - `/api/credentials/categories/${category}`, - ); + try { + const response = await fetch( + `/api/credentials/categories/${encodeURIComponent(category)}`, + ); + if (!response.ok) { + const errorText = await response.text().catch(() => ''); + throw new Error(`HTTP ${response.status}: ${errorText || response.statusText}`); + } + const result = await response.json(); + // ... existing mapping logic below + // (no change) + } catch (error) { + throw this.handleCredentialError(error, `Fetching credentials for category: ${category}`); + }Note: keep the existing mapping block; only wrap/encode and add error handling.
132-141: Encode key and include response details for non-404 errors.Avoid path parsing issues and improve diagnostics; still return empty object for 404 as you do today.
- const response = await fetch(`/api/credentials/${key}`); - if (!response.ok) { - if (response.status === 404) { - // Return empty object if credential not found - return { key, value: undefined }; - } - throw new Error(`Failed to fetch credential: ${key}`); - } - return response.json(); + try { + const response = await fetch(`/api/credentials/${encodeURIComponent(key)}`); + if (!response.ok) { + if (response.status === 404) { + return { key, value: undefined }; + } + const errorText = await response.text().catch(() => ''); + throw new Error(`HTTP ${response.status}: ${errorText || response.statusText}`); + } + return response.json(); + } catch (error) { + throw this.handleCredentialError(error, `Fetching credential '${key}'`); + }
223-234: Encode key in update endpoint path.Protect against special characters in credential keys.
- `/api/credentials/${credential.key}`, + `/api/credentials/${encodeURIComponent(credential.key)}`,
274-279: Encode key in delete endpoint path.Same rationale as update/get.
- const response = await fetch(`/api/credentials/${key}`, { + const response = await fetch(`/api/credentials/${encodeURIComponent(key)}`, { method: "DELETE", });archon-ui-main/src/components/settings/FeaturesSection.tsx (2)
42-44: Avoid null sentinels; prefer throwing and handling in the outer catch.
.catch(() => null)uses null to signal failure. Let the Promise reject and rely on your existing try/catch to set fallbacks, or move this fetch into a small service for consistent error handling.- fetch('/api/projects/health').catch(() => null), + fetch('/api/projects/health'),If you want to keep partial results without failing all requests, wrap only this call in its own try/catch outside Promise.all and set schema defaults in that local catch.
57-63: Gate debug logs or remove for production.Console logs are helpful during development but noisy in production. Consider gating behind a dev flag.
- console.log('🔍 Projects health response:', { + if (import.meta.env.DEV) console.log('🔍 Projects health response:', { response: projectsHealthResponse, ok: projectsHealthResponse?.ok, status: projectsHealthResponse?.status, url: '/api/projects/health' });Apply similarly to the subsequent "Projects health data" log.
archon-ui-main/src/services/mcpClientService.ts (5)
99-101: Remove unused baseUrl property.It’s never referenced; drop it to avoid noUnused* warnings.
- // Always use relative URLs to go through Vite proxy - private baseUrl = ''; + // REST calls use relative URLs to traverse the Vite proxy
92-92: Duplicate import of getApiUrl.There are two identical imports; keep only one.
-import { getApiUrl } from '../config/api';
129-134: Harden error parsing across all endpoints (JSON or text).
await response.json()will throw for non-JSON/error-empty bodies, masking the original status. Add a helper and reuse it for every non-OK branch.Apply this pattern here and to all other non-OK checks below:
- if (!response.ok) { - const error = await response.json(); - throw new Error(error.detail || 'Failed to create MCP client'); - } + if (!response.ok) { + throw new Error(await this.getErrorMessage(response, 'Failed to create MCP client')); + }Add this helper method inside the class:
private async getErrorMessage(res: Response, fallback: string): Promise<string> { try { const data = await res.clone().json(); return data?.detail || data?.message || `HTTP ${res.status} ${res.statusText}` || fallback; } catch { const text = await res.text().catch(() => ''); return text || `HTTP ${res.status} ${res.statusText}` || fallback; } }
110-116: Apply the hardened error parsing consistently.For each endpoint, replace per-endpoint JSON parsing with the shared getErrorMessage helper to avoid parse errors and improve messages.
I can push a quick sweep commit if desired.
Also applies to: 141-149, 155-167, 172-183, 193-203, 209-219, 225-233, 239-251, 261-269, 275-287, 293-301, 307-317
401-429: Minor: derive protocol/host directly to avoid surprises behind proxies.If Archon is served behind a reverse proxy, using window.location preserves the externally visible host/protocol.
Alternative snippet (if you prefer not to touch getApiUrl at all):
const { protocol, hostname } = window.location; const mcpUrl = `${protocol}//${hostname}:${mcpPort}/mcp`;
📜 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 (8)
archon-ui-main/src/components/layouts/MainLayout.tsx(1 hunks)archon-ui-main/src/components/settings/FeaturesSection.tsx(2 hunks)archon-ui-main/src/config/api.ts(1 hunks)archon-ui-main/src/services/bugReportService.ts(1 hunks)archon-ui-main/src/services/credentialsService.ts(7 hunks)archon-ui-main/src/services/mcpClientService.ts(14 hunks)archon-ui-main/src/services/testService.ts(1 hunks)python/src/server/api_routes/knowledge_api.py(1 hunks)
🧰 Additional context used
📓 Path-based instructions (7)
archon-ui-main/**/*.{ts,tsx}
📄 CodeRabbit inference engine (CLAUDE.md)
archon-ui-main/**/*.{ts,tsx}: Never return null to indicate failure in the frontend; throw an Error with details instead
Use database task status values directly in the UI with no mapping: todo, doing, review, done
Files:
archon-ui-main/src/components/layouts/MainLayout.tsxarchon-ui-main/src/services/testService.tsarchon-ui-main/src/components/settings/FeaturesSection.tsxarchon-ui-main/src/services/credentialsService.tsarchon-ui-main/src/services/bugReportService.tsarchon-ui-main/src/config/api.tsarchon-ui-main/src/services/mcpClientService.ts
archon-ui-main/src/components/**
📄 CodeRabbit inference engine (CLAUDE.md)
Place reusable UI components under archon-ui-main/src/components/
Files:
archon-ui-main/src/components/layouts/MainLayout.tsxarchon-ui-main/src/components/settings/FeaturesSection.tsx
archon-ui-main/src/{components,hooks,pages}/**/*.{ts,tsx}
📄 CodeRabbit inference engine (CLAUDE.md)
archon-ui-main/src/{components,hooks,pages}/**/*.{ts,tsx}: State naming: use is[Action]ing for loading states (e.g., isSwitchingProject)
State naming: use [resource]Error for error messages
State naming: use selected[Resource] for current selections
Files:
archon-ui-main/src/components/layouts/MainLayout.tsxarchon-ui-main/src/components/settings/FeaturesSection.tsx
archon-ui-main/src/services/**/*.{ts,tsx}
📄 CodeRabbit inference engine (CLAUDE.md)
archon-ui-main/src/services/**/*.{ts,tsx}: Place API communication and business logic under archon-ui-main/src/services/
Service method naming in frontend should follow: get[Resource]sByProject(projectId) for scoped queries
Service method naming in frontend should follow: getResource for single resource fetch
Service method naming in frontend should follow: createResource for creates
Service method naming in frontend should follow: update[Resource](id, updates) for updates
Service method naming in frontend should follow: deleteResource for soft deletes
Files:
archon-ui-main/src/services/testService.tsarchon-ui-main/src/services/credentialsService.tsarchon-ui-main/src/services/bugReportService.tsarchon-ui-main/src/services/mcpClientService.ts
python/src/**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
python/src/**/*.py: Fail fast on service startup failures (crash with clear error if credentials, database, or any service cannot initialize)
Fail fast on missing configuration or invalid environment settings
Fail fast on database connection failures; do not hide connection issues
Fail fast on authentication/authorization failures; halt the operation and surface the error
Fail fast on data corruption or validation errors; let Pydantic raise
Fail fast when critical dependencies are unavailable (required service down)
Never store invalid data that would corrupt state (e.g., zero embeddings, null foreign keys, malformed JSON); fail instead
For batch processing, complete what you can and log detailed failures per item
Background tasks should finish queues but log failures clearly
Do not crash on a single WebSocket/event failure; log and continue serving other clients
If optional features are disabled, log and skip rather than crashing
External API calls should retry with exponential backoff; then fail with a clear, specific error
When continuing after a failure, skip the failed item entirely; never persist partial or corrupted results
Include context about the attempted operation in error messages
Preserve full stack traces with exc_info=True in Python logging
Use specific exception types; avoid catching generic Exception
Never return None to indicate failure; raise an exception with details
For batch operations, report both success counts and detailed failure lists
Target Python 3.12 and keep line length at 120 characters
Use Ruff for linting (errors, warnings, unused imports, style) and keep code Ruff-clean
Use Mypy for static type checking and keep code type-safe
Enable auto-formatting on save in IDEs to maintain consistent Python style
Files:
python/src/server/api_routes/knowledge_api.py
python/src/server/**
📄 CodeRabbit inference engine (CLAUDE.md)
Keep the main FastAPI application under python/src/server/
Files:
python/src/server/api_routes/knowledge_api.py
python/src/server/api_routes/**/*.py
📄 CodeRabbit inference engine (CLAUDE.md)
Implement API route handlers under python/src/server/api_routes/
Files:
python/src/server/api_routes/knowledge_api.py
🧠 Learnings (2)
📚 Learning: 2025-09-02T19:42:59.116Z
Learnt from: CR
PR: coleam00/Archon#0
File: CLAUDE.md:0-0
Timestamp: 2025-09-02T19:42:59.116Z
Learning: Applies to archon-ui-main/src/services/**/*.{ts,tsx} : Place API communication and business logic under archon-ui-main/src/services/
Applied to files:
archon-ui-main/src/services/mcpClientService.ts
📚 Learning: 2025-09-02T19:42:59.116Z
Learnt from: CR
PR: coleam00/Archon#0
File: CLAUDE.md:0-0
Timestamp: 2025-09-02T19:42:59.116Z
Learning: Applies to archon-ui-main/src/services/**/*.{ts,tsx} : Service method naming in frontend should follow: update[Resource](id, updates) for updates
Applied to files:
archon-ui-main/src/services/mcpClientService.ts
🧬 Code graph analysis (1)
archon-ui-main/src/services/testService.ts (1)
archon-ui-main/src/config/api.ts (1)
API_BASE_URL(54-54)
🔇 Additional comments (4)
archon-ui-main/src/config/api.ts (1)
9-23: The verification script above will locate anyvite.config.tsfiles and display nearby proxy settings to confirm that relative URLs in production are properly reverse-proxied. Once we have that output, we can finalize whether the fallback to''is safe in production or if we need to documentVITE_API_URLusage.archon-ui-main/src/services/credentialsService.ts (2)
59-60: LGTM: forcing proxy via relative paths.Using a blank baseUrl with a clarifying comment ensures all calls traverse the Vite proxy. Good alignment with the PR goal.
250-259: LGTM: consistent error pipeline.createCredential already funnels errors through handleCredentialError; good consistency.
archon-ui-main/src/services/mcpClientService.ts (1)
398-416: Ignore getApiUrl contract concern: getApiUrl() only returns an empty string (for relative URLs) or the VITE_API_URL value—it never returns “/api,” so the URL fallback logic in createArchonClient() remains correct.Likely an incorrect or invalid review comment.
| "ready": True, | ||
| "credentials_loaded": True, | ||
| "schema_valid": True, | ||
| } |
There was a problem hiding this comment.
🛠️ Refactor suggestion
Derive readiness flags instead of hard-coding True.
Hard-coding ready/credentials/schema as True can mislead the UI and mask startup/config issues. Compute them from actual state and env so external clients don’t proceed prematurely.
Apply this diff within the response block:
- "ready": True,
- "credentials_loaded": True,
- "schema_valid": True,
+ "ready": credentials_loaded and schema_status["valid"],
+ "credentials_loaded": credentials_loaded,
+ "schema_valid": True,And add this snippet inside knowledge_health() before building result:
import os # if not already imported at top
credentials_loaded = bool(
os.getenv("SUPABASE_URL")
and (os.getenv("SUPABASE_SERVICE_ROLE_KEY") or os.getenv("SUPABASE_ANON_KEY"))
)
# Optionally include LM provider keys if required for your ready gate:
# credentials_loaded = credentials_loaded and bool(os.getenv("OPENAI_API_KEY") or os.getenv("ANTHROPIC_API_KEY"))Would you like me to wire a lightweight Supabase connectivity probe with a short timeout to further validate readiness?
🤖 Prompt for AI Agents
In python/src/server/api_routes/knowledge_api.py around lines 886 to 889, the
readiness fields are hard-coded to True which can mislead clients; compute
credentials_loaded and ready from environment/state instead: add an import for
os if missing and insert the provided credentials_loaded assignment before
building result, then replace "credentials_loaded": True and "ready": True with
"credentials_loaded": credentials_loaded and "ready": credentials_loaded (or a
combined boolean that also checks any LM provider keys you require); similarly
compute and set "schema_valid" based on your actual schema validation logic
instead of True; optionally consider wiring a short-timeout Supabase
connectivity probe to further validate readiness.
|
Thanks for this @croakingtoad! I will be reviewing PRs this weekend 👍 |
|
FYI - @croakingtoad 's fixes are also in my PR relating the localhost/versus external IP being used incorrectly on the /health call |
|
@croakingtoad - we have cleaned up some things that have caused merge conflicts and potentially resolved the first issue (backend startup errors). Would you be able to test/rebase if needed? |
|
Closed => no response |
…updates (coleam00#565) Bumps the npm_and_yarn group with 1 update in the /CATACLYSM_STUDIOS_INC/PMOVES-PROVISIONS/docker-stacks/jellyfin-ai/api-gateway directory: [qs](https://github.com/ljharb/qs). Bumps the npm_and_yarn group with 1 update in the /pmoves/ui directory: [diff](https://github.com/kpdecker/jsdiff). Updates `qs` from 6.13.0 to 6.14.1 - [Changelog](https://github.com/ljharb/qs/blob/main/CHANGELOG.md) - [Commits](ljharb/qs@v6.13.0...v6.14.1) Updates `diff` from 7.0.0 to 8.0.3 - [Changelog](https://github.com/kpdecker/jsdiff/blob/master/release-notes.md) - [Commits](kpdecker/jsdiff@7.0.0...v8.0.3) --- updated-dependencies: - dependency-name: qs dependency-version: 6.14.1 dependency-type: indirect dependency-group: npm_and_yarn - dependency-name: diff dependency-version: 8.0.3 dependency-type: indirect dependency-group: npm_and_yarn ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
…g empty string (#593) * Fix: warn on unknown \$nodeId.output refs instead of silently returning empty string (#565) Typos in $nodeId.output references (e.g. $clasify.output vs $classify.output) were silently replaced with empty strings with no log output, making them impossible to diagnose. Load-time validation also skipped these refs, so broken workflows loaded without error. Changes: - substituteNodeOutputRefs(): log warn 'dag_node_output_ref_unknown_node' for unknown nodeId - resolveOutputRef(): log warn 'condition_output_ref_unknown_node' for unknown nodeId (distinguishes "node doesn't exist" from "node has empty output") - validateDagStructure(): validate $nodeId.output refs in when: and prompt: fields at load time - Update tests to assert warning is logged for unknown node refs - Move condition-evaluator.test.ts to its own bun test invocation (uses mock.module) Fixes #565 * fix: address review findings from PR #593 - Add loader.test.ts coverage for load-time $nodeId.output ref validation (5 tests: bad when: ref, bad prompt: ref, valid refs, multi-source lastIndex reset, bash: excluded from load-time validation) - Rename _match → match in dag-executor substituteNodeOutputRefs callback - Tighten warn count assertion to .toBe(2) in condition-evaluator.test.ts - Update validateDagStructure JSDoc to list fourth validation check - Update resolveOutputRef JSDoc to document warn-vs-silent distinction - Add inline comment on lastIndex = 0 explaining stateful g-flag regex reset
…g empty string (coleam00#593) * Fix: warn on unknown \$nodeId.output refs instead of silently returning empty string (coleam00#565) Typos in $nodeId.output references (e.g. $clasify.output vs $classify.output) were silently replaced with empty strings with no log output, making them impossible to diagnose. Load-time validation also skipped these refs, so broken workflows loaded without error. Changes: - substituteNodeOutputRefs(): log warn 'dag_node_output_ref_unknown_node' for unknown nodeId - resolveOutputRef(): log warn 'condition_output_ref_unknown_node' for unknown nodeId (distinguishes "node doesn't exist" from "node has empty output") - validateDagStructure(): validate $nodeId.output refs in when: and prompt: fields at load time - Update tests to assert warning is logged for unknown node refs - Move condition-evaluator.test.ts to its own bun test invocation (uses mock.module) Fixes coleam00#565 * fix: address review findings from PR coleam00#593 - Add loader.test.ts coverage for load-time $nodeId.output ref validation (5 tests: bad when: ref, bad prompt: ref, valid refs, multi-source lastIndex reset, bash: excluded from load-time validation) - Rename _match → match in dag-executor substituteNodeOutputRefs callback - Tighten warn count assertion to .toBe(2) in condition-evaluator.test.ts - Update validateDagStructure JSDoc to list fourth validation check - Update resolveOutputRef JSDoc to document warn-vs-silent distinction - Add inline comment on lastIndex = 0 explaining stateful g-flag regex reset
…g empty string (coleam00#593) * Fix: warn on unknown \$nodeId.output refs instead of silently returning empty string (coleam00#565) Typos in $nodeId.output references (e.g. $clasify.output vs $classify.output) were silently replaced with empty strings with no log output, making them impossible to diagnose. Load-time validation also skipped these refs, so broken workflows loaded without error. Changes: - substituteNodeOutputRefs(): log warn 'dag_node_output_ref_unknown_node' for unknown nodeId - resolveOutputRef(): log warn 'condition_output_ref_unknown_node' for unknown nodeId (distinguishes "node doesn't exist" from "node has empty output") - validateDagStructure(): validate $nodeId.output refs in when: and prompt: fields at load time - Update tests to assert warning is logged for unknown node refs - Move condition-evaluator.test.ts to its own bun test invocation (uses mock.module) Fixes coleam00#565 * fix: address review findings from PR coleam00#593 - Add loader.test.ts coverage for load-time $nodeId.output ref validation (5 tests: bad when: ref, bad prompt: ref, valid refs, multi-source lastIndex reset, bash: excluded from load-time validation) - Rename _match → match in dag-executor substituteNodeOutputRefs callback - Tighten warn count assertion to .toBe(2) in condition-evaluator.test.ts - Update validateDagStructure JSDoc to list fourth validation check - Update resolveOutputRef JSDoc to document warn-vs-silent distinction - Add inline comment on lastIndex = 0 explaining stateful g-flag regex reset
4 participants
Summary
This PR fixes critical issues preventing Archon from working correctly when accessed via external IP addresses (not localhost), particularly affecting cloud deployments on DigitalOcean, AWS, etc.
Issues Fixed
Root Cause Analysis
1. Health Endpoint Response Format
The
/api/healthendpoint was missing required fields that the frontend expected:ready: truefield caused frontend to think backend wasn't initializedhealthData.ready === truebut field didn't exist2. API URL Construction Issues
Frontend services were bypassing the Vite proxy:
http://68.183.152.55:8181/api/.../api/...to go through Vite proxy at port 37373. Static URL Configuration
Service classes set
baseUrlonce at instantiation:getApiUrl()which returned the backend port directlyChanges Made
Backend Changes
ready,credentials_loaded, andschema_validfields to health endpoint responseFrontend Changes
getApiUrl()to return empty string in development (forces proxy usage)/api/healthTesting Performed
✅ Backend health endpoint returns correct response format
✅ All API calls go through Vite proxy (port 3737)
✅ Settings page loads without errors
✅ Projects toggle works correctly
✅ No more "Backend Service Startup Failure" popups
✅ Configuration checks pass successfully
Deployment Impact
This fix is critical for:
🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
Bug Fixes
Refactor