Module chat_agents 

Group-chat sub-agent hierarchy — local LLM / Webizen agents are bound to human principals.

Each participant may run a different model/backend. Processed outcomes (not raw prompts) can be shared with other participants when an explicit OutcomeSharingPolicy permits it.

Structs

OutcomeSharingPolicy

Permission for sharing Webizen-processed outcomes (summaries, grounded answers) with peers.

ParticipantAgentConfig

Local user’s sub-agent binding for a chat session.

PeerAgentSummary

Summary of a peer’s disclosed agent for cooperative multi-LLM briefing.

Enums

AgentBackendKind

OutcomeVisibility

Functions

bind_local_roster_agent

Bind the local session projection to a named roster agent immediately before its turn. The roster remains the source of truth; this small session record is used solely to attribute the persisted chat message.

build_cooperative_agents_block

can_relay_agent_outcome

Whether a processed agent outcome may be relayed to group peers.

can_use_in_peer_llm_context

Whether another participant’s sub-agent may cite this outcome in inference context.

can_view_agent_outcome

Whether viewer_did may read a peer’s shared agent outcome in relay / UI.

collect_peer_agent_summaries

Collect peer agent disclosures from committed, shareable agent messages in the session.

compile_sub_agent_did

Deterministic sub-agent DID scoped to principal + session.

decorate_local_agent_message

Apply sub-agent metadata to an agent-role message from the local principal.

default_outcome_sharing

default_outcome_sharing_for_profile

ensure_local_agent_config

fresh_local_agent_config

is_participant

load_local_agent_config

local_agent_display_name

participant_dids

save_local_agent_config

update_outcome_sharing

validate_ingested_agent_message