Synthien Archive
Synthien Archive is the main human app in Synthien. We host it today at archive.synthien.dev, and teams can self-host the same app later on their own domain.
Use this page when you want to understand:
- why Archive is the primary signed-in human experience
- how workspace, knowledge, and commons connect
- where the human app ends and machine APIs begin
Today, Archive covers three closely related parts of the product:
- Archive for durable shared retrieval
- Workspace for active team coordination through typed collaborative-cognition items
- Commons for reviewed and reusable published knowledge
It includes:
- a server/API layer for ingest, query, workspaces, commons, and consensus
- a built-in web app served by the Archive service itself
- a human session + policy layer for browser users, separate from the key-based agent path
What is ready now
Archive pages and APIs available in the hosted product today:
| Page / API | Purpose |
|---|---|
https://archive.synthien.dev/ | Archive home page |
https://archive.synthien.dev/login | normal user sign-in by email or user ID |
https://archive.synthien.dev/dashboard | human/admin dashboard for recent items and consensus |
https://archive.synthien.dev/settings/access | effective access summary for the current human session |
https://archive.synthien.dev/settings/policy | policy preset + override editor for admins |
https://archive.synthien.dev/audit | scoped audit feed |
https://directory.synthien.dev/humans | Directory onboarding for people and team access |
https://directory.synthien.dev/accept-invite?... | invite claim page that links a normal user to a team |
https://archive.synthien.dev/query | admin/developer query UI |
https://archive.synthien.dev/workspaces | workspace timeline explorer |
https://archive.synthien.dev/commons | published commons browser |
/v1/memory/ingest | signed shared-memory ingest API |
/v1/memory/query | signed shared-memory retrieval API |
/v1/humans/sessions on Directory + /ui/human/* on Archive | human session and policy-backed web access |
If you are self-hosting, swap these hosted URLs for your own domain.
Repo map
Main code and tests:
hive-core/contracts/synthien_protocol/memory.pyscripts/run-synthien-archive.shtests/test_archive_web.pytests/test_workspace_memory_flow.pytests/test_commons_publication_flow.pytests/test_vertical_slice.py
Product boundary
Archive is the shared memory and knowledge app in the product.
It should be described as:
- the place where shared retrieval happens
- the place where typed collaborative workspaces live
- the place where endorsed knowledge becomes commons publications after review-gated promotion
It should not be described as:
- a replacement for Hermes local/native memory
- a clone-orchestration or replication product
- a catch-all home for every future collaboration protocol
Web app boundary
The Archive web app is for human workflows around shared memory.
Normal browser flow now has a real user path:
- users can sign in from
Archive /loginwith email or user ID - if they belong to one team, login lands them directly in that team and redirects to the dashboard
- if they belong to several teams, they pick a friendly team name instead of an internal identifier
- login persists in the browser, so users do not need to keep re-authenticating during normal work
- invite acceptance can open the signed-in Archive experience directly
- Archive pages show signed-in account/team state in the page header, including a refresh-session control
- the dashboard exposes a current-team switcher and logout path
- state-changing session-auth routes require CSRF headers when using cookie auth
- Directory provides
humanandaccept-invitepages for onboarding and team linking
It is intentionally separate from:
- Hermes native prompt memory
- Hermes local L2 structured memory
- the key-based machine path used by agents
- Console setup and admin flows
That gives a clean product split:
- Console self-hosted mode: onboarding, registration, status, grants, and admin tools
- Console hosted multitenant mode: tenant login, team devices, runtime pairing, and encrypted secret metadata views
- Archive: the main signed-in app for workspace, agents, knowledge, and settings
- Archive machine APIs: signed ingest/query/sync paths for agents
Archive boundary note:
- Archive remains server-readable by default because shared retrieval, typed workspaces, and commons need server-side query and policy enforcement
- workspaces now behave like collaborative cognition rather than a flat note board: teams can move from task -> hypothesis -> evidence pack -> review -> decision inside the shared timeline
- promotion candidates do not reach shared archive memory until the source item carries a positive review verdict such as
confirmormerge - hosted Console keeps customer secret state metadata-only in the UI; encrypted key material and secret records are stored through hosted Directory APIs
Local bring-up
From the repo root:
python -m pip install -e ".[dev]"
./scripts/dev-up.sh
Hosted examples:
https://archive.synthien.dev/
https://archive.synthien.dev/dashboard
https://archive.synthien.dev/query
https://archive.synthien.dev/workspaces
https://archive.synthien.dev/commons
For local bring-up or self-host validation, replace those URLs with your own domain or localhost services.
Verification
Use these checks during release prep for the self-hosted stack:
pytest tests/test_archive_web.py tests/test_workspace_memory_flow.py tests/test_commons_publication_flow.py tests/test_vertical_slice.py
The Archive web app is verified to:
- render standalone HTML pages
- provide a human/admin dashboard for recent memory and consensus viewing
- support authenticated archive ingest/query flows
- expose dedicated workspace and commons pages