Communication: hbf-nlp

1-hop view of how this service communicates with its siblings. For the full system view, see service-communication.md.

Calls Out To

Service	Protocol	Purpose	Key calls / queues
hbf-core	HTTP (hbf-core-api + direct)	Session, tenant, KB, pipeline management	`ChatSessionsClient.findById/update()`, `TenantsClient.findById()`, `KnowledgeBaseClient.list()`, `KnowledgeBaseArticleClient.findAllByKnowledgeBase()`, GET `/nlp-pipelines/{pipelineId}`, GET `/activities/training-content`, POST/PATCH test set runs, PATCH pipeline
helvia-rag-pipelines	HTTP POST	RAG processing, search, corpus, training	`/pipelines/{id}:process`, `:search`, `/corpus` (PUT), `:train` -- Bearer auth per pipeline
hbf-notifications	HTTP POST	Training notifications (optional)	`/organizations/{orgId}/notifications/`
Dialogflow	gRPC (@google-cloud/dialogflow SDK)	Intent detection	SDK method calls
Google Gemini	HTTP (@google/genai SDK)	LLM completions	SDK method calls
OpenAI	HTTP (openai SDK)	LLM completions, embeddings	SDK method calls
Azure OpenAI	HTTP (custom client)	LLM completions, embeddings	Custom HTTP calls
MySQL	TCP (TypeORM)	Message metadata persistence	MessageMetadata entity
Redis	TCP (keyv/redis)	Optional cache	Key-value cache (optional)

Service	Protocol	How
hbf-bot	HTTP POST	`/tenants/{tenantId}/process` (Bearer token)
hbf-console	HTTP POST	`/organizations/{orgId}/tenants/{tenantId}/process`

POST /organizations/:orgId/tenants/:tenantId/process -- process message (console)
POST /organizations/:orgId/tenants/:tenantId/nlp-pipelines/:pipelineId/process -- process with specific pipeline
POST /tenants/:tenantId/process -- process message (moderator/bot)
POST /organizations/:orgId/tenants/:tenantId/train -- train all pipelines
POST /organizations/:orgId/tenants/:tenantId/nlp-pipelines/:pipelineId/train -- train specific pipeline
POST /organizations/:orgId/tenants/:tenantId/sessions/:sessionId/analyze -- LLM session summary
POST /llm-request -- direct LLM completion (bot only)
POST /detect-language -- language detection (bot only)

hbf-core: Multiple REST calls via hbf-core-api + direct HTTP (see "Calls Out To")
helvia-rag-pipelines: POST /pipelines/{id}:process, :search, PUT /pipelines/{id}/corpus, POST /pipelines/{id}:train -- Bearer auth
hbf-notifications: POST /organizations/{orgId}/notifications/ (optional, for training status)
LLM providers: Dialogflow, Gemini, OpenAI, Azure OpenAI per pipeline configuration