When to use this file:
- MCP server reads this as context for the audio-classifier whenever a recording is dropped in a channel bound to this retreat (
@Funga autoingest audio here for budapest-2026).
- This is the live retreat config. Pre-retreat validation runs out of
retreats/budapest-test/overview.md (a sandbox with the same 21 sessions + 6 synthesis extras) — that config stays alive as a permanent regression-test fixture. Real retreat recordings land here; test recordings stay there.
- Keep
Session ID values stable — the bot uses them as session_id in transcript frontmatter, and changing them mid-retreat fragments the data. Titles + descriptions are free to evolve as the agenda firms up.
Collaborative Budapest Retreat 2026 — Audio Ingest Config
This file tells Funga how to classify audio recordings dropped in the retreat’s Slack channel. The bot uses the Sessions table (and the Synthesis Extras mapping further down) as context for a Haiku classification call against each upload’s caption + transcript head, then files the transcript to _transcripts/budapest-2026/<date>-<short-id>.md with the matched session + team in frontmatter.
Curated synthesis lives under retreats/budapest-2026/synthesis/<session>/ — five universal category files (Convergence, Divergence, Uncertainties & Risks, Impact Stories, Proposals & Opportunities), one Themes file, optional Gap Analysis when enabled, plus one file per session-specific Synthesis Extra (see table at the bottom).
Slack channel
#budapest_26_recordings (C0B19P75U2Y). To bind:
@Funga autoingest audio here for budapest-2026
Bot will confirm. After binding, drop audio with a caption that hints at session + group — e.g. “Trio book round — Atossa+Mille+Indy” or “Bioregions table sensemaking, host Sandrine”.
How classification works
- Participant uploads an audio file to
#budapest_26_recordings with a short caption.
- Bot transcribes (Whisper) → calls Haiku with this whole file + caption + first 800 chars of transcript → gets back
{session_id, team_label, participants, confidence, reasoning}.
- Always files the transcript best-effort. High-confidence (>70%) classifications file silently with a confirmation. Low-confidence ones post “reply if I got this wrong” in the thread; user replies trigger reclassification.
The matching is semantic, not regex — vague, paraphrased, or partial captions are fine as long as they carry enough signal. Even an empty caption works if the transcript content matches a session description.
For cross-cutting sessions (context-setting, closing-reflection, village-news), recordings aggregate across all 5 days. A caption like “day 2 village news” or “Wednesday closing reflection” is helpful — the day signal helps reviewers (it’s surfaced as the recording’s recorded_at) but doesn’t fragment the session.
Sessions
| Session ID | Title | Date | Time | Description |
|---|
| context-setting | Context setting & agenda overview | Cross-cutting | varies | Daily morning framing — agenda walkthrough, why we’re here, what today builds toward. Recordings aggregate across all 5 days. Use captions like “Day 2 context setting” or “Wednesday morning agenda”. |
| day1-requisite-diversity | Day 1 — Requisite diversity mapping | 2026-05-11 | 15:35 | 9-cell grid sensemaking — System Layers, Ways of Working, Stakeholder Proximity. Physical placement → cross-dimension sensemaking → bridge into brings/needs. Look for where clusters form vs where gaps stand out. |
| day2-symbios-deep-dive | Day 2 — SymbiOS Deep Dive (presentation) | 2026-05-12 | 10:00 | Origins of SymbiOS, reef metaphor, 3 components at altitude (Reef DNA / Solidarity, Reef Metabolism / Portfolio-Learning, Reef Skeleton / Governance), solidarity deep dive. Plenary presentation with discussion. |
| day2-os-reflective-dialogue | Day 2 — OS Reflective Dialogue | 2026-05-12 | 11:15 | OS Gallery Walk + annotation + facilitated reflective dialogue. Stations: values charter, ways of working, mutualism, value exchange, requisite diversity, relational health. Includes post-lunch synthesis/feedback recordings continuing the dialogue. Move values from page into body. |
| day2-living-values | Day 2 — Living our values | 2026-05-12 | 14:40 | Fishbowl theatre with actors + observers. Scenarios played out. Rehearsal of OS-aligned behaviours — what does it feel like when a value is being upheld vs eroded. |
| day2-mutualism-practice | Day 2 — Mutualism in Practice / Needs and Offers | 2026-05-12 | 16:15 | Putting OS into practice — system give/get, solidarity infrastructure, what shows up across teams. Brings / needs / how we want to show up. |
| day3-sensemaking | Day 3 — Sensemaking & Areas for Exploration | 2026-05-13 | 10:46 | Self-facilitated small groups orient to lines of enquiry. Scene-setting from seed person, round of inputs from each participant’s standpoint, reframe enquiry, identify where opportunities lie. Group-named team labels (e.g. “bioregions table”, “finance group”, “Mind the Gap”). Includes Mind-the-Gap groups that continued into Day 4 morning. |
| day3-areas-exploration | Day 3 — DELETE | 2026-05-13 | 17:46 | FUTURE / looking forward — enriching portfolio logics, surfacing domains/positions, argument building based on who we are and what we’ve sensed. Combines small-group exploration + the long shareback session. Inquiry areas and domains where we might act. |
| day4-portfolio-frames-activations | Day 4 — Portfolio Frames & Activation Ideas | 2026-05-14 | varies | Combined Day 4 portfolio work: frame selection, co-design ideation mashups (sandwich groups, cross-org mashups), portfolio shape exploration, world-fair-style sharebacks, tipping points, business models, intergenerational collab, mother earth, ecology of possibilities, opt-out OS, digital infrastructure. One unified bucket spanning the day’s morning ideation through afternoon sharebacks. |
| day5-open-space | Day 5 — Open Space | 2026-05-15 | 14:35 | Open Action Space — self-organized workspaces. Recordings include both the small-group breakouts and the consolidated shareback. The room ahead — what we are doing together, independently of funding. |
| day5-fundraising | Day 5 — Fundraising Session | 2026-05-15 | varies | Dedicated fundraising conversations — coordination across rooms, follow-up sessions. Distinct from the broader Open Space because it ran as its own focused track. |
Notes:
- Cross-cutting rows’ Date column says “Cross-cutting” so the bot understands the same session_id is reused across multiple days.
- Day-specific rows use ISO dates so the bot can correlate “Wednesday” / “Tuesday” / “Day 2” caption hints back to the right session.
- If the live agenda shifts (timing changes, sessions get renamed, new sessions appear), edit titles/descriptions in place — keep
Session ID stable.
Some sessions get session-specific synthesis sections in addition to the 5 universal categories (Convergence, Divergence, Uncertainties & Risks, Impact Stories, Proposals & Opportunities). Each row below adds one extra card to that session’s synthesis output. The synthesis pipeline reads this table, asks Sonnet for the extra section by name, and writes it to retreats/budapest-2026/synthesis/<session>/<kebab-name>.md. The frontend renders an additional card; projector mode cycles it right after the universals.
| Session ID | Extra Section Name |
|---|
| day1-requisite-diversity | Gaps and Surprises in the System |
| day2-os-reflective-dialogue | First Reactions and Reflections |
| day3-sensemaking | What Only We Could Do |
| day4-portfolio-frames-activations | Portfolio Logics and Activation Ideas |
| day5-open-space | Commitments Independent of Funding |
Group labels — emergent, not predefined
Groups at SCAC retreats self-organize and often re-form between sessions, so we don’t predefine a fixed enum. For each recording, the bot extracts a free-form team_label from the caption + transcript content + uploader. Examples:
"Atossa+Mille+Indy" (uploader-named participants)
"the bioregions table" (caption-described group)
"Joaquin's group, day 2 morning" (caption with topical anchor)
- (blank) when there’s no team — e.g. plenary presentations or solo intros
Cross-group synthesis is grouped by session_id, not by team. “Across the recordings in day3-sensemaking-1” rather than “what team alpha said vs team beta.”
Caption conventions for participants
No strict format — semantic matching handles vague captions. Examples that all work:
- “Atossa, Mille and Indy on bioregional finance” — names as group identity, topic for session
- “morning trio book round” — descriptive group + session inference
- “Day 2 village news” — day signal helps disambiguate cross-cutting sessions
- “finance discussion” — topic only; team_label inferred from uploader or transcript speakers
- (empty caption) — bot falls back to transcript content + uploader for both fields
If the bot misclassifies, anyone in the thread can reply with the correction and the bot will re-file:
> ✅ Filed to _transcripts/budapest-2026/2026-05-12-abc123.md
> Session: Day 1 — Requisite diversity mapping · Team: Bioregions · Confidence: 64%
> If I got this wrong, reply in this thread with more context.
[user] actually this was the day-2 reflective dialogue
[bot] 🔁 Reclassified to Session: Day 2 — OS Reflective Dialogue · Confidence: 91%
retreats/README.md — index of retreat configs
retreats/budapest-test/overview.md — pre-retreat test sandbox with the same sessions; stays alive as a permanent Funga regression fixture
retreats/budapest-2026/<other>.md — curated synthesis docs (gap analysis, session summaries) sit alongside this overview
retreats/budapest-2026/synthesis/<session>/ — auto-generated synthesis output per session (Funga writes here)
_transcripts/budapest-2026/ — raw transcripts written here by the bot