Skip to content

overview glossary

khanhthanhdev edited this page Jun 1, 2026 · 1 revision

Glossary

Project-specific terms and abbreviations used throughout the Infi codebase.

Term Definition
ACP Agent Client Protocol — a stdio-based protocol for communicating with coding agents. Infi uses it to spawn and orchestrate external agents like Codex or Claude Code.
Analysis A single research session. Contains a user query, an intent classification, one or more runs, and a structured report. Persisted in SQLite.
Analysis Block A typed piece of an analysis report — can be a metric, stance, projection, methodology note, counter-thesis, or other structured artifact.
Artifact Kind The type discriminator for structured artifacts: metric_snapshot, stance, projection, counter_thesis, methodology_note, decision_criterion, uncertainty, source, metric_explanation.
Final Stance The agent's concluding position on the research question. Submitted via the submit_final_stance MCP tool. Always includes the research disclaimer.
Freshness A system that tracks how old each metric in a stance is. Metrics older than source_freshness_days (default 7) are flagged as stale.
MCP Server The infi-analysis server that runs as a child process of the ACP agent. It exposes tools like fetch_data, submit_metric, submit_stance, and submit_projection that the agent calls to interact with Infi.
Portfolio A collection of holdings or transactions imported from CSV. Can be analyzed to produce allocation review, risk assessment, and expected return models.
Provider A data source adapter (e.g., Finnhub, Yahoo Finance, Polygon). Each provider implements the SourceProvider trait and may require an API key stored in the OS keychain.
Projection A forward-looking scenario analysis with bull/base/bear cases, each containing price targets, time horizons, and probability weights.
Research Plan A structured plan the agent submits at the start of a run, listing the steps it will take to answer the research question.
Run A single execution attempt of an analysis. An analysis can have multiple runs (re-runs). Each run has its own status, progress events, and report snapshot.
RunContext Runtime context for a single analysis run, carrying the analysis ID, run ID, enabled sources, and cancellation token.
Stance The agent's position on a specific aspect of the research. Can be bullish, bearish, or neutral, with supporting evidence.
Structured Artifact A generic typed block in the report that doesn't fit the specific metric/stance/projection categories. Has a kind, title, content, and optional columns/rows.
Tauri Channel A streaming IPC mechanism used to push progress events from the Rust backend to the frontend in real time during a run.
Verification Status The freshness status of a stance metric: fresh, stale, or unverifiable. Computed by comparing the metric's timestamp against the configured freshness window.

Clone this wiki locally