Documentation for the SEMOSS core runtime, agents and harnesses, engine integrations, and deployment. Start with the introduction and backend architecture, or choose the path below that matches your task.
| Task | Guide |
|---|---|
| Understand the platform and repositories | Introduction to SEMOSS |
| Follow a request through the backend | Backend architecture |
| Run the platform locally | Semoss Docker examples |
| Run local backend changes | Java developer onboarding and Monolith local Docker examples |
| Develop the web experience | semoss-ui |
| Configure Kubernetes or deployment properties | SEMOSS-deployment |
- Agents overview: agents, workspaces, rooms, runs, tools, and skills.
- The SEMOSS harness: model/tool loop, prompt composition, budgets, compaction, hooks, and delegation.
- Agent configuration: create a workspace, select models and execution targets, attach tools and skills, and configure limits.
- Run lifecycle and streaming: durable states, asynchronous submission, approvals, cancellation, event polling, and cluster boundaries.
- Agent skills: project-backed skills, authoring, discovery, attachment, staging, and loading.
- Workbench default agents: frontend defaults, saved conversations, and deployment of system agents.
- PowerPoint visual inspection: the specialized presentation workflow and validation.
- Claude Code integration: the alternate
claude_codeharness. - AskRoom: persistent model turns with caller-managed tool continuation.
| Concept | Guide |
|---|---|
| Pixel language | Syntax and execution |
| Reactors | Reactor framework |
| Engines | Engine abstraction |
| Data and queries | DataFrames and QueryStructs |
| Execution context | The Insight object |
| Projects, agents, and skill packages | Project types |
- Database engines
- Model engines
- Vector engines
- Storage engines
- Function engines
- Projects
- Supporting engine Docker examples
- Internal databases and agent persistence
- Authentication and authorization
- Java–Python communication
- Monolith integration and API request flow
- Monolith endpoints by application:
/apiengine routes, compatibility APIs, webhooks, health, tokens, and administrator setup. - Monolith webhooks and change notifications: GitHub pushes, Microsoft mailbox/calendar subscriptions, callbacks, and delivery diagnostics.
- Monolith README
- Development guides
- Configuration and environment
- Java developer onboarding
- How-to index
- Write a custom reactor
- Interact with databases
- Work with frames
- Use the NounStore
- Advanced Pixel scripting
- TypeSafe / Jev models
The GenAI client index covers provider clients, embeddings, tokenization, model limits, and generation. See also reasoning and effort and the provider pages for OpenAI, Anthropic, Google, Bedrock, and Text Generation Inference.
The Python GAAS tools index documents Python access to SEMOSS models, databases, vectors, storage, functions, and server communication. These clients are integration components; the native Java agent harness is documented separately under agents.
- Local Docker setup
- Local Docker configuration
- Supporting engine examples
- Cloud and cluster architecture
- Central asset storage
- Redis and ZooKeeper synchronization
- CI/CD workflows
- Kubernetes, infrastructure, and semoss-artifacts properties
Use the source links and the configuration shipped with your image to resolve version-specific behavior. Pixel examples containing <...> are templates: replace the placeholders with accessible resources and run them in an authenticated SEMOSS context.
When changing a reactor, harness, project type, or deployment setting, update the relevant guide and this index. Keep examples aligned with the same revisions of Semoss, Monolith, semoss-ui, and the packaged project assets.