Pioneer documentation
Pioneer is a local-first engineering workspace for conversations, evidence, repeatable workflows, and human-controlled execution. Pioneer Flow combines deterministic processing, model reasoning, native scripts, saved definitions, and durable run history in one visual workspace.
These pages document the current Pioneer repository build. They distinguish what is configured, what Pioneer has directly observed, what actually executed, what was persisted, what a person authorized, and what remains an inference.
Start here
-
Run Pioneer
Start the Windows-first development build and confirm that the backend, frontend, and selected AI provider are the ones you intended.
-
Understand Pioneer Flow
Learn how Canvas drafts, saved Flows, executions, evidence, and repairs relate without collapsing into one hidden action.
-
Read the trust model
Understand Pioneer's evidence classes and the authority boundaries around proposals, changes, saves, and runs.
-
Diagnose a problem
Work from symptoms to evidence, then make the smallest justified change.
Product surfaces
| Surface | Use it for |
|---|---|
| Chat | Conversation, engineering work, Flow inspection, and explicit proposals or actions. |
| Canvas | Building and inspecting the current visual Flow draft. |
| Executions | Live and durable run outcome, stages, results, approvals, and run evidence. |
| Models | Provider checks, local model inventory, model lifecycle, defaults, and context evidence. |
| Machine | Observed hardware, runtime, provider reachability, profile identity, and storage state. |
| Memory | Durable records and explicit relevance-based Recall. |
| Sessions | Persisted conversation, tool, approval, and replay evidence. |
| Logs | Application and Spine activity for operational diagnosis. |
| Settings | Endpoints, credentials, preferences, tools, and provider configuration. |
Complete a real workflow
Start with Your First Chat and Your First Flow. When a run fails, use the task-oriented Diagnose, Repair, and Verify guide without collapsing diagnosis, apply, rerun, and verification into an autonomous loop.
Documentation scope
This is the canonical user and operator documentation tree. Product plans, benchmark reports, implementation journals, demo runbooks, and model-facing skills remain internal repository material and are not published here.