Exodus is the open-source migration, evidence, and sovereignty platform for measuring, planning, and executing movement out of closed vendor control surfaces.

It is not just a one-time export tool.

It is a governed product capability with five connected functions:

1. Discover provider topology, assets, dependencies, control surfaces, and confidence gaps.
2. Plan ingestion, archival, processing, and budget waves before moving bulk data.
3. Preserve and ingest raw evidence, manifests, and validation signals.
4. Process and normalize documents, media, metadata, and derived analytical records.
5. Score and guide progress and the next highest-value steps to reduce control-surface dependence.

python3 scripts/validate_examples.py && ./scripts/publish_to_github.sh SocioProphet exodus public

• Program / repo / product name: Exodus
• User-facing UI: Exodus Dashboard
• Core score: Exit Readiness Index (ERI)
• Secondary score: Provider Control Surface (PCS)
• Execution plane: Exodus Engine

Most migration tooling is file-centric and event-blind. It can copy objects but cannot answer:

• How dependent are we, really?
• What is still trapped behind provider-specific APIs, identities, formats, workflows, and AI services?
• What is the next highest-value move to reduce dependence?
• Can we prove progress with evidence rather than anecdotes?
• What should we ingest now, what should stay metadata-only, and what should be archived later?

Exodus answers those questions.

Providers in scope first:

• Apple
• Google
• Microsoft

Possible later expansion:

• Dropbox
• Slack
• Notion
• Adobe
• GitHub
• Zoom
• Salesforce
• Atlassian

• Phase 0 — scope, trust boundaries, target open states
• Phase 1 — crawl, topology, index, metadata, size/temperature, exportability
• Phase 2 — ingestion, archival, processing, and budget proposal
• Phase 3 — preservation and intake execution
• Phase 4 — normalization, extraction, token/entity/timeline processing
• Phase 5 — dashboard governance, recommendations, continuous monitoring

• Open execution primitives: Kafka + Beam + Drill
• Canonical storage model: raw lake + derived lake + immutable manifests
• Dashboard role: map, plan, prove, score, guide, and report
• Source adapters: export/archive tools, API collectors, local trusted-device collectors, notes/media/document adapters

• apps/console/ — Exodus Dashboard
• packages/contracts/ — canonical contracts, schemas, and typed object grammar
• packages/scoring/ — ERI, PCS, recommendation scoring, and explanation logic
• packages/adapters/ — provider adapters and collector capability models
• schemas/ — persisted-record JSON Schemas
• docs/ — product, architecture, phase specs, domain, roadmap
• examples/ — synthetic tenants and worked examples
• backlog/ — epics and milestone backlog

• License: MIT
• Open-first formats: JSONL, Parquet, Iceberg
• Open execution: Beam on Flink/Spark, Drill query plane
• Cloud use is allowed as infrastructure, but open primitives remain canonical

Deliver a working dashboard that can:

1. ingest Apple, Google, and Microsoft metadata and export state,
2. build a provider topology snapshot and asset census,
3. compute an explainable initial ERI and PCS,
4. surface blockers and recommendations,
5. produce a machine-readable Phase 2 budget proposal.

Exodus now includes an explicit Phase 3 / Phase 4 document pipeline inspired by IBM's staged unstructured-data processing model, but implemented on open primitives. See docs/processing/sire-inspired-pipeline-2026.md and docs/processing/h-net-dynamic-chunking-evaluation.md.

Exodus now treats DocumentIR, deterministic canonical chunking, enrichment outputs, and benchmark runs as first-class objects. The first benchmark lane compares the deterministic baseline against an H-Net adaptive overlay on the same normalized corpus.