Test Endpoint

---
description: Test all v1 API endpoints against live Vercel deployment. Reads tests/endpoints.yaml, runs each test, logs results per endpoint, inspects Vercel logs on failure, loops until all pass.
---

# Test Endpoint

Test all TradeGuru v1 API endpoints against the live deployment.

## Variables

ENDPOINTS_FILE: tests/endpoints.yaml
LOG_DIR: tests/logs/endpoint
BASE_URL: https://tradeguru.com.au
VERCEL_PROJECT_DIR: /mnt/c/Users/jehma/Desktop/TradeGuru/expo-chatgpt-clone

## Instructions

### Step 1: Read Configuration

Read `$ENDPOINTS_FILE` to load all endpoint definitions.

### Step 2: Register Test Device

Before running tests, register a fresh test device:

```bash
curl -s -X POST $BASE_URL/api/v1/device/register \
  -H "Content-Type: application/json" \
  -d '{"platform":"ios","os_version":"18.0","app_version":"1.0.0","locale":"en-AU","timezone":"Australia/Sydney"}'

Extract the device_id from the response. This device_id is used for all tests with auth: device.

If registration fails, STOP and report the error. All subsequent tests depend on a valid device.

Step 3: Run All Endpoint Tests

For EACH endpoint in the YAML file, execute the test:

Build the curl command based on endpoint config:

• auth: none — No auth headers
• auth: device — Add -H "X-Device-ID: {device_id}" and inject device_id into body if body has device_id field
• auth: device_override — Use the device_id field from the endpoint config (for testing invalid IDs)
• method — Use -X {method}
• body — If not null, serialize as JSON with -d '{...}'
• content_type — Default application/json unless specified otherwise

For SSE streaming endpoints (expect.content_type: text/event-stream):

• Use curl -N --max-time 30 to capture the stream
• Look for event: block and event: done in output
• If event: error appears, capture the error data

Check the response against expectations:

• expect.status — HTTP status code must match
• expect.body_contains — Response body must contain this string
• expect.content_type — Response Content-Type header must match
• expect.content_type_contains — Content-Type must contain this substring
• expect.sse_events — For SSE, these event types must appear

Record result as PASS or FAIL.

Step 4: Write Individual Log Files

For EACH endpoint test, write a markdown log file at $LOG_DIR/{slug}.md:

# Endpoint Test: {slug}

**Tested:** {ISO timestamp}
**URL:** {method} {base_url}{path}
**Auth:** {auth type}
**Status:** PASS / FAIL

## Request

```bash
{exact curl command used}

Response

HTTP Status: {status code} Content-Type: {content-type header}

{response body - first 2000 chars}

Expectations

Check	Expected	Actual	Result
Status	{expected}	{actual}	PASS/FAIL
Body contains	{expected}	{found?}	PASS/FAIL

Vercel Logs (if FAIL)

{vercel log output for this endpoint}

Diagnosis (if FAIL)

{AI analysis of what went wrong and suggested fix}

### Step 5: Inspect Vercel Logs on Failure

For ANY test that FAILs with a 500 error:

```bash
cd $VERCEL_PROJECT_DIR && npx vercel logs --limit 20 2>&1

Capture the relevant log entries and include them in the endpoint's log file.

Then READ the source file for the failing endpoint to diagnose the issue.

Step 6: Fix and Re-test Loop

If ANY test FAILs:

1. Analyze the failure using Vercel logs + source code
2. Identify the root cause
3. Fix the source file in the Expo project at $VERCEL_PROJECT_DIR
4. Before saving Edge Runtime files, check for Node.js-only APIs (Buffer, require('crypto'), fs, path, etc.)
5. Run cd $VERCEL_PROJECT_DIR && npx tsc --noEmit to verify compilation
6. Commit and push the fix: cd $VERCEL_PROJECT_DIR && git add -A && git commit -m "fix: {description}" && git push origin main
7. Wait 30-60 seconds for Vercel to deploy
8. Re-run ONLY the failing tests
9. Repeat until all tests PASS

Maximum iterations: 10. If tests still fail after 10 fix cycles, stop and report all remaining failures with full diagnostics.

Step 7: Final Summary

After all tests pass (or max iterations reached), print a summary table:

============================================================
 ENDPOINT TEST RESULTS
============================================================

 Total: {N} endpoints
 Passed: {P}
 Failed: {F}
 Fix iterations: {I}

 Results:
 {PASS/FAIL icon} {slug} — {status code} {notes}
 ...

 Log files: tests/logs/endpoint/*.md
============================================================

Important Rules

• ALWAYS register a fresh device before testing — don't reuse stale device IDs
• For streaming endpoints (chat, vision), use --max-time 30 to prevent hanging
• Include -s (silent) flag on curl for cleaner output parsing
• Use -w "\n%{http_code}" to reliably extract HTTP status codes
• Never modify existing endpoints (assistants-, stripe-, user-*) while fixing v1 bugs
• After fixing Edge Runtime files, always verify no Node.js-only APIs are used (Buffer, require('crypto'), fs, path, __dirname)
• Log files are append-safe — each run overwrites the slug's file with latest results