Video analysis

.gitignore

@@ -4,4 +4,8 @@
 *.swp
 tags
 node_modules
-__pycache__
+__pycache__
+*scratch*
+*.vscode
+*/private/*
+*test_consumer_payload.py

deploy/systemd/rag-service-consumer.service

@@ -0,0 +1,24 @@
+[Unit]
+Description=RAG Service RabbitMQ Consumer for rag.evalix.xyz
+Wants=network-online.target
+After=network-online.target
+
+[Service]
+Type=simple
+User=frappe
+Group=frappe
+WorkingDirectory=/home/frappe/frappe-bench
+Environment=PYTHONUNBUFFERED=1
+Environment=PATH=/home/frappe/frappe-bench/env/bin:/usr/local/bin:/usr/bin:/bin
+ExecStart=/usr/local/bin/bench --site rag.evalix.xyz execute rag_service.scripts.console_consumer.start_consumer
+Restart=always
+RestartSec=10
+StartLimitIntervalSec=0
+KillSignal=SIGINT
+TimeoutStopSec=30
+StandardOutput=journal
+StandardError=journal
+SyslogIdentifier=rag-service-consumer
+
+[Install]
+WantedBy=multi-user.target

private/local_docker_setup.md

@@ -0,0 +1,112 @@
+# Local Docker setup for rag_service
+
+This guide creates a fresh local Frappe site named `rag_service.localhost` using Docker, Postgres, Redis, and the local `rag_service` app source from this repository.
+
+The recommended path is the root stack script:
+
+```sh
+cd /Users/TAP/Documents/Git/LMS
+cp .env.example env.local
+./scripts/start_local_docker.sh
+```
+
+## External RabbitMQ
+
+RabbitMQ is created externally in CloudAMQP:
+
+```text
+https://customer.cloudamqp.com/login
+```
+
+Fill these root `env.local` values from the CloudAMQP instance:
+
+```dotenv
+RABBITMQ_HOST=
+RABBITMQ_PORT=5672
+RABBITMQ_VIRTUAL_HOST=
+RABBITMQ_USERNAME=
+RABBITMQ_PASSWORD=
+RABBITMQ_PLAGIARISM_RESULTS_QUEUE=plagiarism_feedback
+RABBITMQ_FEEDBACK_RESULTS_QUEUE=feedback_results
+```
+
+## Required local DocTypes discovered in code
+
+The active folders reviewed were:
+
+- `rag_service/core`
+- `rag_service/feedback_utils`
+- `rag_service/utils`
+- `rag_service/rag_service/doctype`
+
+The local setup needs these settings available:
+
+- `RabbitMQ Settings`: used by `utils/rabbitmq_consumer.py` and `utils/queue_manager.py`.
+- `GCS Settings`: used by `utils/gcp_service_client.py`.
+- `RAG Settings`: used by `core/assignment_context_manager.py` to fetch assignment and student context from CRM.
+- `LLM Settings`: used by most feedback generation paths.
+- `Gemini Settings`: used by Gemini-specific image evaluation paths.
+- `Prompt Segment` and `Prompt Template`: required by template-driven feedback generation.
+
+The root script symlinks `/home/frappe/frappe-bench/apps/rag_service` to `/workspace/rag_service`, installs `rag_service` in editable mode, installs `business_theme_v14`, and seeds `RabbitMQ Settings`, `GCS Settings`, and `RAG Settings` from `env.local`. It also creates optional `LLM Settings` and `Gemini Settings` rows when the related env values are present.
+
+## Values to set before use
+
+In root `env.local`, set:
+
+```dotenv
+RAG_TAP_LMS_BASE_URL=http://tap_lms.localhost:8000
+RAG_ASSIGNMENT_CONTEXT_ENDPOINT=/api/method/tap_lms.api.get_assignment_context
+RAG_STUDENT_CONTEXT_ENDPOINT=/api/method/tap_lms.api.get_student_context
+RAG_TAP_LMS_API_KEY=
+RAG_TAP_LMS_API_SECRET=
+RAG_ENABLE_CACHING=1
+RAG_CACHE_DURATION_DAYS=1
+```
+
+For model-backed feedback generation, also set one of:
+
+```dotenv
+RAG_LLM_PROVIDER=Gemini
+RAG_LLM_MODEL_NAME=
+RAG_LLM_API_KEY=
+RAG_LLM_API_SECRET=
+```
+
+or:
+
+```dotenv
+RAG_GEMINI_MODEL_NAME=
+RAG_GEMINI_PROJECT_ID=
+RAG_GEMINI_LOCATION=us-central1
+RAG_GEMINI_CREDENTIALS_JSON={}
+```
+
+## Start RAG after setup
+
+```sh
+docker compose --env-file env.local -f docker/local/docker-compose.yml exec rag-dev bash -lc "cd /home/frappe/frappe-bench && bench start"
+```
+
+Open:
+
+```text
+http://rag_service.localhost:8001
+```
+
+Login:
+
+```text
+User: Administrator
+Password: the RAG_ADMIN_PASSWORD value from `env.local`
+```
+
+## Consumer
+
+The app exposes a Frappe command for the RabbitMQ consumer. After the site is installed and RabbitMQ settings are valid, run it from the RAG dev container:
+
+```sh
+docker compose --env-file env.local -f docker/local/docker-compose.yml exec rag-dev bash -lc "cd /home/frappe/frappe-bench && bench --site rag_service.localhost start-rag-consumer"
+```
+
+If the command path changes, inspect `rag_service/hooks.py` and `rag_service/commands`.

private/local_setup.py

@@ -0,0 +1,67 @@
+import os
+
+import frappe
+
+
+def _as_float(value, default):
+    try:
+        return float(value)
+    except (TypeError, ValueError):
+        return default
+
+
+def _as_int(value, default):
+    try:
+        return int(value)
+    except (TypeError, ValueError):
+        return default
+
+
+def seed_optional_records():
+    """Seed optional local LLM records from environment variables.
+
+    Single DocTypes are set from the shell script. These regular DocTypes need
+    insert/update logic and are only useful when credentials are present.
+    """
+    llm_model_name = os.getenv("RAG_LLM_MODEL_NAME")
+    llm_api_key = os.getenv("RAG_LLM_API_KEY")
+    gemini_model_name = os.getenv("RAG_GEMINI_MODEL_NAME")
+    gemini_project_id = os.getenv("RAG_GEMINI_PROJECT_ID")
+
+    if llm_model_name and llm_api_key:
+        name = f"Local {os.getenv('RAG_LLM_PROVIDER', 'Gemini')} {llm_model_name}"
+        if frappe.db.exists("LLM Settings", name):
+            doc = frappe.get_doc("LLM Settings", name)
+        else:
+            doc = frappe.new_doc("LLM Settings")
+            doc.name = name
+
+        doc.provider = os.getenv("RAG_LLM_PROVIDER", "Gemini")
+        doc.model_name = llm_model_name
+        doc.api_key = llm_api_key
+        doc.api_secret = os.getenv("RAG_LLM_API_SECRET", "")
+        doc.temperature = _as_float(os.getenv("RAG_LLM_TEMPERATURE"), 0)
+        doc.max_tokens = _as_int(os.getenv("RAG_LLM_MAX_TOKENS"), 2000)
+        doc.is_active = 1
+        doc.is_default = 1
+        doc.save(ignore_permissions=True)
+
+    if gemini_model_name and gemini_project_id:
+        name = f"Local Gemini {gemini_model_name}"
+        if frappe.db.exists("Gemini Settings", name):
+            doc = frappe.get_doc("Gemini Settings", name)
+        else:
+            doc = frappe.new_doc("Gemini Settings")
+            doc.name = name
+
+        doc.model_name = gemini_model_name
+        doc.project_id = gemini_project_id
+        doc.location = os.getenv("RAG_GEMINI_LOCATION", "us-central1")
+        doc.credentials_json = os.getenv("RAG_GEMINI_CREDENTIALS_JSON", "{}")
+        doc.temperature = _as_float(os.getenv("RAG_LLM_TEMPERATURE"), 0)
+        doc.max_tokens = _as_int(os.getenv("RAG_LLM_MAX_TOKENS"), 2000)
+        doc.is_active = 1
+        doc.is_default = 1
+        doc.save(ignore_permissions=True)
+
+    frappe.db.commit()