How synthwave-solutions/hermes-agent works
Hermes Agent, the engine behind the SynthPulse Agentic Workstation, is a multi-surface AI agent: users converse through CLI/TUI or messaging platforms, the gateway and conversation runtime coordinate model calls and tools, and the agent returns responses while retaining session context. Its broader product includes learning and recall, scheduled automation, delegation, and multiple execution backends. The README establishes these product capabilities; source excerpts provide only partial implementation evidence, so uncertain internal links are omitted rather than inferred.
Main entry points: conversation_loop.py Conversation loop · gateway Messaging channels · console_engine.py CLI and TUI · server.py Gateway service
Components 20
User surfaces
- CLI and TUI
hermes_cli/console_engine.py - Messaging channels
gateway - Gateway service
tui_gateway/server.py - ACP server
acp_adapter/server.py
Agent runtime
- Conversation loop
agent/conversation_loop.py - Context compression
agent/context_compressor.py - Model metadata
agent/model_metadata.py - Data redaction
agent/redact.py - Localization
agent/i18n.py
Agent capabilities
- Subagents
agent/delegation_provider.py - Learning and skills
agent/learning_graph.py - Session search
agent - Cron automation
cron
Models and tools
- Model providers
agent/provider_registry.py - Tool execution
agent/tool_executor.py - Execution backends
tools/environments
Session and memory
- Session state
agent/session_persistence.py - Agent memory
agent/memory_manager.py - Conversation archive
agent/conversation_compression.py
Other components
- User
Connections 21
- User → CLI and TUI: starts chat
README.md - User → Messaging channels: sends messages
README.md - CLI and TUI → Conversation loop: starts conversation
- Messaging channels → Gateway service: routes messages
README.md - Gateway service → Conversation loop: runs agent
tui_gateway/server.py - ACP server → Conversation loop: exposes agent
acp_adapter/server.py - Conversation loop → Model providers: calls models
- Conversation loop → Tool execution: dispatches tools
- Tool execution → Execution backends: uses environments
README.md - Conversation loop → Context compression: compresses context
agent/conversation_loop.py - Conversation loop → Session state: persists turns
- Conversation loop → Model metadata: uses model data
agent/conversation_loop.py - Conversation loop → Data redaction: sanitizes content
agent/conversation_loop.py - Conversation loop → Localization: localizes responses
- Conversation loop → Subagents: spawns subagents
README.md - Conversation loop → Learning and skills: learns from tasks
README.md - Conversation loop → Session search: searches history
README.md - Cron automation → Messaging channels: delivers automation
README.md - Learning and skills → Agent memory: curates knowledge
README.md - Context compression → Conversation archive: archives context
- Conversation loop → User: returns response