[RHIDP-11652] Add BYOM Architecture for Lightspeed (#200)

Author: Jdubrick

.github/workflows/test.yml

@@ -48,12 +48,8 @@ jobs:
             cliArgs: "-f compose.yaml -f compose-dynamic-plugins-root.yaml"
           - name: "orchestrator-workflow"
             cliArgs: "-f compose.yaml -f orchestrator/compose.yaml"
-          - name: "developer-lightspeed-minimal"
-            cliArgs: "-f compose.yaml -f developer-lightspeed/compose.yaml"
           - name: "developer-lightspeed"
-            cliArgs: "-f compose.yaml -f developer-lightspeed/compose-with-ollama.yaml"
-          - name: "developer-lightspeed-safety-guard"
-            cliArgs: "-f compose.yaml -f developer-lightspeed/compose-with-ollama.yaml -f developer-lightspeed/compose-with-safety-guard-ollama.yaml"
+            cliArgs: "-f compose.yaml -f developer-lightspeed/compose.yaml"
 
     name: "${{ matrix.tool }} compose - ${{ matrix.composeConfig.name }}${{ matrix.os != 'ubuntu-24.04' && format(' - {0}', matrix.os) || '' }}${{ matrix.userConfig == 'true' && ' - user config' || '' }}"
     runs-on: ${{ matrix.os }}

.gitignore

@@ -22,10 +22,9 @@ compose.override.yaml*
 !/configs/extra-files/github-app-credentials.example.yaml
 /developer-lightspeed/configs/extra-files/*
 !/developer-lightspeed/configs/extra-files/lightspeed-stack.yaml
-!/developer-lightspeed/configs/extra-files/run.yaml
-!/developer-lightspeed/configs/extra-files/run-no-guard.yaml
-!/developer-lightspeed/configs/extra-files/templates/
-!/developer-lightspeed/configs/extra-files/templates/placeholder.json
+!/developer-lightspeed/configs/extra-files/config.yaml
+!/developer-lightspeed/configs/extra-files/rhdh-profile.py
+!/developer-lightspeed/configs/extra-files/templates
 
 
 # Kept for backward compatibility, but now Git-ignored

default.env

@@ -23,6 +23,11 @@ CATALOG_ENTITIES_EXTRACT_DIR=/extensions
 #AUTH_GITHUB_CLIENT_ID=
 #AUTH_GITHUB_CLIENT_SECRET=
 
+# Max entry size for dynamic plugin installation (in bytes)
+# Note: Lightspeed needs >= 40000000
+# and Orchestrator needs >= 30000000
+MAX_ENTRY_SIZE=40000000
+
 # Node environment for extensions plugin installation
 NODE_ENV=development
 
@@ -52,67 +57,71 @@ SEGMENT_WRITE_KEY=gGVM6sYRK0D0ndVX22BOtS7NRcxPej8t
 # Developer Lightspeed Configuration
 # ==============================================================================
 # For detailed documentation, see: developer-lightspeed/README.md
-# Note: Ollama is pre-configured by default - no setup needed. External providers require configuration below.
 
 # ------------------------------------------------------------------------------
-# Ollama Provider Configuration (Default - Pre-configured)
+# Inference Provider Enablement
 # ------------------------------------------------------------------------------
-# Pre-configured and ready to use - typically no changes needed
-OLLAMA_URL=http://ollama:11434
-# Change to use different model
-OLLAMA_MODEL=llama3.2:1b
-# Optional: mount local models directory
-# OLLAMA_MODELS_PATH=/path/to/.ollama
+## Set to 'true' to enable models from that provider. I.e. ENABLE_VLLM=true
+## Leave blank to disable that provider. I.e. ENABLE_VLLM=
+## Note: You can have multiple enabled. I.e. ENABLE_VLLM=true ENABLE_OPENAI=true
+ENABLE_VLLM=
+ENABLE_VERTEX_AI=
+ENABLE_OPENAI=
+ENABLE_OLLAMA=
+
 
 # ------------------------------------------------------------------------------
-# External Provider Enablement
+# Ollama Provider Configuration
+# Required if ENABLE_OLLAMA=true
 # ------------------------------------------------------------------------------
-# Set ONE to 'true' if using external provider instead of Ollama
-# ENABLE_VLLM=
-# ENABLE_OPENAI=
-# ENABLE_VERTEX_AI=
+## Must end with /v1
+OLLAMA_URL=
+
 
 # ------------------------------------------------------------------------------
 # vLLM Provider Configuration
-# ------------------------------------------------------------------------------
 # Required if ENABLE_VLLM=true
-# Must end with /v1
+# ------------------------------------------------------------------------------
+## Must end with /v1
 VLLM_URL=
-# Leave empty if no auth required
+## Leave empty if no auth required
 VLLM_API_KEY=
-# Optional: max tokens (default: 4096)
+## Optional: max tokens (default: 4096)
 # VLLM_MAX_TOKENS=4096
-# Optional: TLS verification (default: true)
+## Optional: TLS verification (default: true)
 # VLLM_TLS_VERIFY=true
 
+
 # ------------------------------------------------------------------------------
 # OpenAI Provider Configuration
-# ------------------------------------------------------------------------------
 # Required if ENABLE_OPENAI=true
-# Get from https://platform.openai.com/api-keys
+# ------------------------------------------------------------------------------
+## Get from https://platform.openai.com/api-keys
 OPENAI_API_KEY=
 
+
 # ------------------------------------------------------------------------------
 # Vertex AI Provider Configuration (Experimental)
-# ------------------------------------------------------------------------------
 # Required if ENABLE_VERTEX_AI=true
-# Absolute path to GCP service account JSON
+# ------------------------------------------------------------------------------
+## Absolute path to GCP service account JSON on the host
 VERTEX_AI_CREDENTIALS_PATH=
-# Your GCP project ID
+## Your GCP project ID
 VERTEX_AI_PROJECT=
-# Optional: GCP region (default: us-central1)
+## Optional: GCP region (default: us-central1)
 # VERTEX_AI_LOCATION=us-central1
 
+
 # ------------------------------------------------------------------------------
-# Llama Guard Settings
+# Validation Settings
 # ------------------------------------------------------------------------------
-# The safety provider uses inline::llama-guard, so SAFETY_MODEL must be a
-# llama-guard variant (e.g. llama-guard3:1b).
-# Defaults to llama-guard3:1b if not set.
-SAFETY_MODEL=
-# SAFETY_URL is auto-configured by the compose files.
-# You should not need to set this manually.
-SAFETY_URL=
-# Only required if connecting to a remote safety endpoint that requires auth.
-# Not needed when using the local safety-ollama container (the default).
-SAFETY_API_KEY=
+## Enables query validation.
+## Will limit queries to RHDH specific requests.
+ENABLE_VALIDATION=
+## Required if ENABLE_VALIDATION=true
+## Must be one of the enabled providers. I.e. if ENABLE_OPENAI=true then VALIDATION_PROVIDER=openai
+VALIDATION_PROVIDER=
+## Required if ENABLE_VALIDATION=true
+## Must be an available model for an enabled provider.
+## I.e. ENABLE_OPENAI=true VALIDATION_MODEL_NAME=gpt-4o-mini
+VALIDATION_MODEL_NAME=