Documentation Index
# Documentation Index
Audience: everyone; this page routes you and defines the vocabulary. Purpose: one-line descriptions of every doc, the core glossary, and reading paths. Prerequisites: none.
# The doc set
| Doc | Description | Audience |
|---|---|---|
| ../README.md | Project pitch, features, install, quick start | Everyone |
| ARCHITECTURE.md | System shape, enforced layering rules, boot and turn lifecycle, state locations, decision records | Framework developers |
| GETTING_STARTED.md | Prerequisites, build, workspace init, model configuration, environment variables, troubleshooting | New users and developers |
| RELURPISH.md | Daily use of the terminal UI: tabs, commands, approvals, postures, state inspection | Users of the coding agent |
| RECIPE_DSL.md | Authoring thoughtrecipes: triggers, tool scopes, steps, captures, families | Agent and workflow authors |
| PROMPT_DSL.md | Prompt file format, front matter, the prompt registry, overrides | Agent and workflow authors |
| CONFIGURATION.md | The relurpify_cfg/ tree: workspace settings, providers, profiles, security policies, tool manifests, validation rules |
Users and framework developers |
| ../CONTRIBUTING.md | DCO sign-off, architecture gates, coverage floor, commit style, release process | Contributors |
| TESTING.md | Test layers, agent suites, execution profiles, tapes, conformance, coverage | Contributors |
| ../CHANGELOG.md | Release history (Keep a Changelog) | Everyone |
| CODE_OF_CONDUCT.md | Community rules and enforcement contact | Everyone |
# Reading paths
- “I want to use the coding agent”: GETTING_STARTED then RELURPISH .
- “I want to author workflows”: RECIPE_DSL and PROMPT_DSL .
- “I want to work on the framework”: ARCHITECTURE , then CONTRIBUTING and TESTING .
- “Something broke”: GETTING_STARTED troubleshooting , then RELURPISH .
# Glossary
These terms are used with exactly these meanings in every doc.
| Term | Meaning |
|---|---|
| Relurpify | This framework; the repository. |
| relurpish | The terminal UI binary; also runs doctor and status. |
| Euclo | The built-in coding agent that relurpish runs. |
| workspace | The project directory the agent works in; the agent acts only inside it. |
relurpify_cfg/ |
The workspace configuration tree; materialized once by doctor, then owned by you. |
.relurpify_state/ |
The runtime state directory: logs, telemetry, audit chain, jobs, memory, artifacts. Output only, never source. |
| capability | Anything the model may do; declared by a descriptor with trust class, effect classes, and schemas. |
| sandbox | Container-isolated command execution; gVisor or Docker backend. |
| HITL | Human-in-the-loop: an approval request that pauses a run for your decision. |
| thoughtrecipe | A declarative .erpe workflow that Euclo routes to and executes; “recipe” for short. |
| intent family | One of nine keyword-classified instruction groups Euclo routes on (debug, review, …). |
| paradigm | A reusable reasoning strategy (ReAct, planner, blackboard, ReWOO, …) that recipes can delegate to. |
| knowledge chunk | The memory atom: content plus provenance, freshness, and typed links into the graph store. |
| context compiler | The deterministic stage that assembles what the model sees each turn under a token budget. |
| envelope | The per-task state carrier that flows through the agent graph. |
Secondary terms are defined on first use: doctor (readiness check), relurplint (workspace linter), dev-agent (agent test CLI), relurpify-runner (background job runner), tape (LLM record/replay fixture), agenttest (YAML-driven end-to-end suite), selection posture (deterministic or model_assisted intent disambiguation).