Configuration Reference

Complete reference for both connectors, the environment variables / system properties, and the default values. For task-oriented guidance, see Configuring the Agent, Tools, RAG, Chat Memory, and Audit Trail.

AI Agent connector — input parameters

connectorId = cibseven-ai-agent. Required: agentName, message. All others optional.

Input group

Parameter	Type	Default	Purpose
agentName	String	— (required)	Logical agent name.
message	Text	— (required)	The user message; may reference process variables.
instruction	Text	empty → bundled default prompt	System prompt text.
instructionMode	Dropdown	replace	How instruction combines with the default: replace / append / prepend. Ignored when instruction empty; unknown value → error.

Agent group

Parameter	Type	Default	Purpose
baseUrl	String	OPENAI_BASE_URL env → https://api.openai.com/v1	OpenAI-compatible endpoint (chat model).
model	String	…defaultModel prop → CIBSEVEN_CONNECT_AI_AGENT_DEFAULT_MODEL env → gpt-5.4-nano	LLM model identifier.
apiKey	String	OPENAI_API_KEY env	Provider API key.
customHeaders	Text (JSON)	{}	Extra HTTP headers (object of string→string) on every request.
reasoningEffort	String	unset	Reasoning effort hint (e.g. low/medium/high), forwarded as-is.
reasoningSummary	String	unset	Enables reasoning summary (auto/concise/detailed); switches to the OpenAI Responses API.
agentDescription	Text	empty	Human-readable description.
persistChatLog	Dropdown	empty = global default	Per-activity override of the chat-log audit variable: ${true} / ${false} / empty. See Audit Trail.

Tools group

See Tools.

Parameter	Type	Default	Purpose
mcpServers	Text (JSON array)	[]	MCP servers: [{"name","url","headers"}]. name (1–32 of [a-zA-Z0-9_-]) prefixes tool names as <name>__<tool>; absent → slug from host/port.
toolClasses	Text (CSV)	empty	Comma-separated FQNs of @Tool classes (public no-arg constructor). Built-in: org.cibseven.connect.ai.agent.impl.ProcessStarterTool.

Memory group

See Chat Memory.

Parameter	Type	Default	Purpose
useChatMemory	Dropdown	${false}	Activate per-conversation memory.
memoryId	String	new UUID (returned as output)	Conversation id to reuse across invocations. Shown only when useChatMemory = ${true}.
chatMemoryMaxMessages	String (int)	20	Sliding-window size (messages retained). Shown only when useChatMemory = ${true}.

RAG group

See RAG.

Parameter	Type	Default	Purpose
pgHost	String	empty (setting it activates RAG)	PostgreSQL host of the pgvector store.
pgPort	String (int)	5432	PostgreSQL port.
pgDatabase	String	empty	Database name.
pgUser	String	empty	Database user.
pgPassword	String	empty	Database password.
pgTable	String	langchain4j_embeddings	pgvector table name.
maxRagResults	String (int)	5	Max documents retrieved per query.
minRagScore	String (double)	0.0	Min cosine-similarity (0.0–1.0); 0.0 = no filtering.
embeddingDimension	String (int)	384	Embedding vector dimension; must match ingestion.
embeddingModelName	String	empty → local AllMiniLmL6V2	Remote embedding model name (uses apiKey); empty → local model.

Output parameters

Source expression	Template variable	Type	Purpose
${output}	agentOutput	String	Final agent text answer (required mapping).
${outputAiMeta}	agentOutput_aiMeta	Map	AI-generated marker.
${memoryId}	memoryId	String	Chat-memory id (when memory active).

Knowledge Ingestor connector

connectorId = cibseven-knowledge-ingestor. Required: content, pgHost. Embeds text into the pgvector store for later RAG retrieval. See RAG.

Inputs

Parameter	Type	Default	Purpose
content	Text	— (required)	Text to embed and store.
source	String	empty	Source id (file/URL/process id), stored as segment metadata.
metadata	Text	empty	Extra metadata as comma-separated key=value pairs.
chunkSize	String (int)	500	Max characters per chunk.
chunkOverlap	String (int)	50	Character overlap between chunks.
embeddingModelName	String	empty → local AllMiniLmL6V2	Remote embedding model name; empty → local.
apiKey	String	OPENAI_API_KEY env	API key for a remote embedding model.
embeddingDimension	String (int)	384	Vector dimension; must match the model.
pgHost	String	— (required)	PostgreSQL host.
pgPort	String (int)	5432	PostgreSQL port.
pgDatabase	String	empty	Database name.
pgUser	String	empty	Database user.
pgPassword	String	empty	Database password.
pgTable	String	langchain4j_embeddings	pgvector table name.

Output

Source expression	Template variable	Type	Purpose
${chunksIngested}	ingestedChunks	Integer	Number of chunks embedded and stored.

Environment variables & system properties

Name	Kind	Default	Effect
OPENAI_API_KEY	env	—	Fallback API key when apiKey is not set.
OPENAI_BASE_URL	env	https://api.openai.com/v1	Fallback chat base URL when baseUrl is not set.
cibseven.connect.ai-agent.defaultModel	system property	—	Deployment-wide default model (wins over the env var).
CIBSEVEN_CONNECT_AI_AGENT_DEFAULT_MODEL	env	—	Deployment-wide default model (fallback for the property).
cibseven.connect.ai-agent.redactContent	system property	false	When true, message/response content in the audit log is replaced by a SHA-256 hash + length.
CIBSEVEN_CONNECT_AI_AGENT_REDACT_CONTENT	env	false	Env fallback for redaction.
cibseven.connect.ai-agent.chatLogVariable.enabled	system property	true	When false, the per-activity chat-log variable is not written (in-memory events still emitted to SLF4J / history).
CIBSEVEN_CONNECT_AI_AGENT_CHAT_LOG_VARIABLE_ENABLED	env	true	Env fallback for the chat-log toggle.
AI_AGENT_ENABLED	env	distro-dependent	Distribution-level master switch for the overlay (run/run4/tomcat/wildfly Docker). See Installation.

Default values at a glance

Setting	Default	Constant
Model (built-in fallback)	gpt-5.4-nano	DEFAULT_MODEL
Chat base URL	https://api.openai.com/v1	DEFAULT_BASE_URL
PostgreSQL port	5432	DEFAULT_PG_PORT
pgvector table	langchain4j_embeddings	DEFAULT_PG_TABLE
Max RAG results	5	DEFAULT_MAX_RAG_RESULTS
Min RAG score	0.0	DEFAULT_MIN_RAG_SCORE
Embedding dimension	384 (AllMiniLmL6V2)	DEFAULT_EMBEDDING_DIMENSION
Chunk size	500	DEFAULT_CHUNK_SIZE
Chunk overlap	50	DEFAULT_CHUNK_OVERLAP
Chat memory window	20	DEFAULT_CHAT_MEMORY_MAX_MESSAGES
Instruction mode	replace	DEFAULT_INSTRUCTION_MODE

CLI: KnowledgeIngestor arguments

See RAG for usage.

Argument	Required	Default
--file	yes	—
--pgHost	yes	—
--pgUser	yes	—
--pgPassword	yes	—
--pgPort	no	5432
--pgDatabase	no	postgres
--pgTable	no	langchain4j_embeddings
--chunkSize	no	500
--chunkOverlap	no	50