Workflows

Research Workflow

Research in Akita follows a cycle of curiosity, investigation, and synthesis:

1. Signal detection — briefings and signal reports surface interesting topics from RSS feeds and web sources
2. Deep research — tracks__request_deep_research produces a comprehensive dossier in 2–4 minutes
3. Follow-up — chain research with parentId to go deeper on specific angles
4. Corpus integration — dossiers are indexed and become searchable context for future work

Collaborative Note-Writing with Claude via MCP

The MCP tools turn Claude into a research collaborator with access to the full knowledge base:

• Search context: corpus__search to find relevant passages across all indexed content
• Read highlights: readwise__get_reading_activity to pull in recent annotations
• Review journal: thoughts__list_thoughts for personal reflections on a topic
• Create content: garden__create_note, garden__update_note to write directly into the garden

A typical session: ask Claude to research a topic, it searches the corpus for relevant book passages and highlights, synthesizes them with web research, and drafts a note — all within a single conversation.

Reading Material Pipeline

Books

Add to Calibre → extract_chapters → summarize_chapters → indexed in corpus
                                                        → create_fiche (personal notes)
                                                        → promote to book entry (review)

Articles

Save to Readwise Reader → highlights synced via API
                        → create_article_entry (from URL, auto-fetches OG metadata)
                        → or create_fiche for extended notes

Deep Research

request_deep_research → dossier generated (2-4 min)
                      → publish_content to blog/essays
                      → or keep as reference in dossier archive

Drafts and Fiches Accumulation

Not everything needs to be published immediately. The system supports gradual accumulation:

• Fiches — dev-only sidecar notes that collect quotes, reactions, and observations about a resource over time. Visible only in development mode.
• Draft notes — hidden from the production site but visible in dev. Ideas can gestate here indefinitely.
• Inbox thoughts — quick captures that can be reviewed and promoted to notes later.

This accumulation is intentional: it mirrors how thinking actually works. An idea encountered in a book today might connect with a podcast heard next month. The fiche is where these connections gather before crystallizing into a published review or essay.

Publication Pipeline

Content moves from private to public through deliberate stages:

Thought / highlight / capture
  → Inbox (raw, unprocessed)
  → Fiche or Draft note (accumulating, shaping)
  → Resource entry or Note (draft: true, visible in dev)
  → Published (draft: false, live on candide.me)
  → Optionally: private (encrypted, requires password)

Each transition is a conscious choice. Nothing is published accidentally. The toggle_draft and toggle_private tools (and dev toolbar buttons) make these transitions frictionless.

Continuity Between Research and Publication

The system is designed so that research flows naturally into publication:

• A deep research dossier can be published directly to the blog via publish_content(dossierId=...)
• A fiche, after accumulating enough material, can be promoted to a full resource entry with one command
• Notes written collaboratively with Claude already live in the garden — they just need draft: false

This continuity means there’s no “export step” or “copy-paste phase.” The content is already where it needs to be; publication is just a visibility toggle.