root/fusion
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs(FN-4223): complete Step 2 — disambiguate research guide wording

Browse Source 
Fusion-Task-Id: FN-4223
Fusion-Task-Lineage: bc114ff2-ee0f-4e96-b1d2-6ba3112a48e4
This commit is contained in:
Fusion
2026-05-13 04:57:23 -07:00
committed by gsxdsm
parent 4885a288bd
commit f2fb27ca23
1 changed files with 11 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
docs/research.md
View File

@@ -2,10 +2,12 @@
[← Docs index](./README.md)
Fusion Research lets you create bounded research runs that search, fetch, and synthesize information with built-in web tools by default, then turn structured findings into actionable tasks — all from the dashboard, CLI, or agent sessions.
Fusion Research is a cited search-and-synthesis pipeline: create bounded research runs, gather sources with built-in web tools, and synthesize findings into actionable task context from the dashboard, CLI, or agent sessions.
---
> **Not pi-autoresearch:** Fusion Research is a cited search/synthesis pipeline. The autonomous try-measure-keep-revert loop from upstream `pi-autoresearch` is a separate domain (experiment sessions); see [naming-decision-2026-05.md](./research/naming-decision-2026-05.md).
## Overview
A **research run** is a persisted workflow that moves through orchestration phases:
@@ -160,7 +162,7 @@ Each finding has two task-facing actions:
## CLI Usage
The `fn research` command provides full research run management from the terminal.
The `fn research` command provides cited-research run management from the terminal (search/fetch/synthesis runs, not experiment sessions).
### Commands
@@ -284,11 +286,11 @@ Additionally, planning-mode sessions and the LLM synthesis agent can opt into ru
| Tool | Description |
|---|---|
| `fn_research_run` | Start a bounded research run. Parameters: `query`, `wait_for_completion`, `max_wait_ms` |
| `fn_research_list` | List recent runs. Parameters: `status`, `limit` |
| `fn_research_get` | Get a run's structured findings. Parameters: `id` |
| `fn_research_cancel` | Cancel an active run. Parameters: `id` |
| `fn_research_retry` | Retry a failed/timed-out run when retryable. Parameters: `id` |
| `fn_research_run` | Start a bounded cited-research run (search/fetch/synthesis). Parameters: `query`, `wait_for_completion`, `max_wait_ms` |
| `fn_research_list` | List recent cited-research runs. Parameters: `status`, `limit` |
| `fn_research_get` | Get a cited-research run's structured findings. Parameters: `id` |
| `fn_research_cancel` | Cancel an active cited-research run. Parameters: `id` |
| `fn_research_retry` | Retry a failed/timed-out cited-research run when retryable. Parameters: `id` |
### Tool responses
@@ -296,6 +298,8 @@ All tools return:
- **Text content** — concise human-readable summary
- **Structured details** — machine-readable metadata (`runId`, `status`, `summary`, `findings`, `citations`, `error`, `setup`)
`ResearchFinding.confidence` in this subsystem is synthesis-emitted confidence from the LLM provider (0–1), not a statistical confidence score.
### Availability checks
Before creating runs, `fn_research_run` checks: