egullickson/motovaultpro
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: Gemini engine module and configuration (#129) #142

New Issue
Closed
opened 2026-02-11 03:49:44 +00:00 by egullickson · 1 comment
egullickson commented 2026-02-11 03:49:44 +00:00
Owner
Copy Link

Relates to #129

Milestone 4: Gemini Engine Module and Configuration

Files

  • ocr/app/engines/gemini_engine.py (NEW)
  • ocr/app/config.py
  • ocr/requirements.txt
  • docker-compose.yml
  • docker-compose.staging.yml
  • docker-compose.prod.yml

Requirements

  • Create GeminiEngine class (standalone, NOT extending OcrEngine -- different interface: OcrEngine.recognize() accepts image bytes and returns text+confidence; GeminiEngine.extract_maintenance() accepts PDF bytes and returns structured JSON)
  • GeminiEngine uses Vertex AI SDK: GenerativeModel("gemini-2.5-flash") with generate_content()
  • GeminiEngine.extract_maintenance(pdf_bytes) validates PDF size BEFORE processing. Reject if raw bytes > 20MB with clear error message including file size. Vertex AI SDK handles base64 encoding internally.
  • Use response_mime_type="application/json" and response_schema for guaranteed JSON structure
  • Gemini response schema uses camelCase field names (serviceName, intervalMiles, intervalMonths, details) to match backend API convention
  • Use prompt from issue #129 for maintenance schedule extraction
  • Response schema enforces maintenanceSchedule[] with serviceName (required), intervalMiles (nullable), intervalMonths (nullable), details (nullable)
  • GeminiEngine._get_client() wraps Vertex AI client initialization in try/except. Catch all exceptions, log full traceback, raise EngineUnavailableError with message: "Vertex AI authentication failed: {exc}"
  • Gemini reads GOOGLE_APPLICATION_CREDENTIALS env var (same as Vision API). Vertex AI SDK ADC supports external_account (WIF) credential type -- no code changes needed, environment setup only
  • Add config settings: VERTEX_AI_PROJECT (required), VERTEX_AI_LOCATION (default: us-central1), GEMINI_MODEL (default: gemini-2.5-flash)
  • Add google-cloud-aiplatform>=1.40.0 to requirements.txt
  • Add environment variables to all docker-compose files for OCR service

Acceptance Criteria

  • GeminiEngine.extract_maintenance(pdf_bytes) returns structured JSON with maintenanceSchedule array
  • Each schedule item has serviceName (required), intervalMiles (nullable), intervalMonths (nullable), details (nullable)
  • PDFs > 20MB (raw bytes) rejected with clear error message including file size
  • WIF authentication failure raises EngineUnavailableError with diagnostic message
  • Configuration reads from environment variables with sensible defaults
  • All field names in response use camelCase

Tests

  • Test files: ocr/tests/test_gemini_engine.py (NEW)
  • Test type: unit (mock Vertex AI SDK)
  • Scenarios:
    • Normal: Valid PDF returns structured maintenance schedules with camelCase fields
    • Edge: PDF with no maintenance content returns empty array
    • Error: PDF > 20MB rejected with size error message
    • Error: Vertex AI authentication failure raises EngineUnavailableError with diagnostic info
    • Error: Gemini API call failure handled gracefully
Relates to #129 ## Milestone 4: Gemini Engine Module and Configuration ### Files - `ocr/app/engines/gemini_engine.py` (NEW) - `ocr/app/config.py` - `ocr/requirements.txt` - `docker-compose.yml` - `docker-compose.staging.yml` - `docker-compose.prod.yml` ### Requirements - Create `GeminiEngine` class (standalone, NOT extending OcrEngine -- different interface: OcrEngine.recognize() accepts image bytes and returns text+confidence; GeminiEngine.extract_maintenance() accepts PDF bytes and returns structured JSON) - GeminiEngine uses Vertex AI SDK: `GenerativeModel("gemini-2.5-flash")` with `generate_content()` - `GeminiEngine.extract_maintenance(pdf_bytes)` validates PDF size BEFORE processing. Reject if raw bytes > 20MB with clear error message including file size. Vertex AI SDK handles base64 encoding internally. - Use `response_mime_type="application/json"` and `response_schema` for guaranteed JSON structure - Gemini response schema uses camelCase field names (serviceName, intervalMiles, intervalMonths, details) to match backend API convention - Use prompt from issue #129 for maintenance schedule extraction - Response schema enforces `maintenanceSchedule[]` with `serviceName` (required), `intervalMiles` (nullable), `intervalMonths` (nullable), `details` (nullable) - `GeminiEngine._get_client()` wraps Vertex AI client initialization in try/except. Catch all exceptions, log full traceback, raise `EngineUnavailableError` with message: "Vertex AI authentication failed: {exc}" - Gemini reads `GOOGLE_APPLICATION_CREDENTIALS` env var (same as Vision API). Vertex AI SDK ADC supports `external_account` (WIF) credential type -- no code changes needed, environment setup only - Add config settings: `VERTEX_AI_PROJECT` (required), `VERTEX_AI_LOCATION` (default: us-central1), `GEMINI_MODEL` (default: gemini-2.5-flash) - Add `google-cloud-aiplatform>=1.40.0` to requirements.txt - Add environment variables to all docker-compose files for OCR service ### Acceptance Criteria - `GeminiEngine.extract_maintenance(pdf_bytes)` returns structured JSON with `maintenanceSchedule` array - Each schedule item has serviceName (required), intervalMiles (nullable), intervalMonths (nullable), details (nullable) - PDFs > 20MB (raw bytes) rejected with clear error message including file size - WIF authentication failure raises `EngineUnavailableError` with diagnostic message - Configuration reads from environment variables with sensible defaults - All field names in response use camelCase ### Tests - **Test files**: `ocr/tests/test_gemini_engine.py` (NEW) - **Test type**: unit (mock Vertex AI SDK) - **Scenarios**: - Normal: Valid PDF returns structured maintenance schedules with camelCase fields - Edge: PDF with no maintenance content returns empty array - Error: PDF > 20MB rejected with size error message - Error: Vertex AI authentication failure raises EngineUnavailableError with diagnostic info - Error: Gemini API call failure handled gracefully
egullickson added the
status
backlog
type
feature
 labels 2026-02-11 03:51:15 +00:00
egullickson added
status
in-progress
 and removed
status
backlog
 labels 2026-02-11 20:31:02 +00:00
egullickson commented 2026-02-11 20:35:24 +00:00
Author
Owner
Copy Link

Milestone: Gemini Engine Module and Configuration

Phase: Execution | Agent: Developer | Status: PASS

Summary

Audited existing GeminiEngine implementation (committed under refs #133) against all #142 acceptance criteria. Found one gap and fixed it.

Acceptance Criteria Verification

# Requirement Status
1 GeminiEngine class, standalone (NOT extending OcrEngine) PASS
2 Vertex AI SDK GenerativeModel with generate_content() PASS
3 PDF size validation before processing, >20MB rejected with file size in error PASS
4 response_mime_type="application/json" and response_schema enforcement PASS
5 camelCase field names (serviceName, intervalMiles, intervalMonths, details) PASS
6 Extraction prompt matches #129 spec PASS
7 Response schema: maintenanceSchedule[] with serviceName required, nullable fields PASS
8 _get_model() wraps init with traceback logging, raises GeminiUnavailableError PASS (fixed)
9 Reads GOOGLE_APPLICATION_CREDENTIALS env var for WIF auth PASS
10 Config: VERTEX_AI_PROJECT, VERTEX_AI_LOCATION (default us-central1), GEMINI_MODEL (default gemini-2.5-flash) PASS
11 google-cloud-aiplatform>=1.40.0 in requirements.txt PASS
12 Environment variables in all 3 docker-compose files PASS

Gap Fixed

  • Added logger.exception() calls for full traceback logging in _get_model() exception handlers
  • Updated catch-all error message to "Vertex AI authentication failed: {exc}" per spec

Test Results

All 18 Gemini engine tests pass:

  • 3 exception hierarchy tests
  • 4 data type tests
  • 2 PDF size validation tests
  • 3 successful extraction tests
  • 4 error handling tests
  • 2 lazy initialization tests

Commit

f9a650a feat: add traceback logging and spec-aligned error message to GeminiEngine (refs #142)

Verdict: PASS | Next: Close issue or continue with parent #129 milestones

## Milestone: Gemini Engine Module and Configuration **Phase**: Execution | **Agent**: Developer | **Status**: PASS ### Summary Audited existing GeminiEngine implementation (committed under refs #133) against all #142 acceptance criteria. Found one gap and fixed it. ### Acceptance Criteria Verification | # | Requirement | Status | |---|---|---| | 1 | GeminiEngine class, standalone (NOT extending OcrEngine) | PASS | | 2 | Vertex AI SDK `GenerativeModel` with `generate_content()` | PASS | | 3 | PDF size validation before processing, >20MB rejected with file size in error | PASS | | 4 | `response_mime_type="application/json"` and `response_schema` enforcement | PASS | | 5 | camelCase field names (serviceName, intervalMiles, intervalMonths, details) | PASS | | 6 | Extraction prompt matches #129 spec | PASS | | 7 | Response schema: `maintenanceSchedule[]` with serviceName required, nullable fields | PASS | | 8 | `_get_model()` wraps init with traceback logging, raises GeminiUnavailableError | PASS (fixed) | | 9 | Reads `GOOGLE_APPLICATION_CREDENTIALS` env var for WIF auth | PASS | | 10 | Config: VERTEX_AI_PROJECT, VERTEX_AI_LOCATION (default us-central1), GEMINI_MODEL (default gemini-2.5-flash) | PASS | | 11 | `google-cloud-aiplatform>=1.40.0` in requirements.txt | PASS | | 12 | Environment variables in all 3 docker-compose files | PASS | ### Gap Fixed - Added `logger.exception()` calls for full traceback logging in `_get_model()` exception handlers - Updated catch-all error message to "Vertex AI authentication failed: {exc}" per spec ### Test Results All 18 Gemini engine tests pass: - 3 exception hierarchy tests - 4 data type tests - 2 PDF size validation tests - 3 successful extraction tests - 4 error handling tests - 2 lazy initialization tests ### Commit `f9a650a` feat: add traceback logging and spec-aligned error message to GeminiEngine (refs #142) *Verdict*: PASS | *Next*: Close issue or continue with parent #129 milestones
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
status
backlog
status/blocked
status
done
status
in-progress
status
ready
status
review
type
bug
type
chore
type
docs
type
feature
No Label
status
in-progress
type
feature
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
egullickson
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: egullickson/motovaultpro#142
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?