Agent Data Hub v0.1.7
Agent Data Hub v0.1.7
Agent Data Hub v0.1.7 is a small Hub View app-readiness release for the
local-first technical preview.
It does not add a new memory model, a hosted service, or autonomous agent
execution. The release makes the visible agent handoff easier to find and easier
to trust: users can see how reviewed ADH context is handed to a chatbot or local
agent, and what setup state ADH can actually verify.
What Changed
Release Hygiene
pyproject.tomlnow matches the public patch version.- The release was checked with a short first-run observation before tagging.
Project Actions In Hub View
- Project pages now show a compact Project actions strip near the top.
- The first action is Hand context to an agent, which jumps directly to the
visible context-pack flow. - The Create context pack action is styled as the primary next step, with a
visible task field instead of a quiet inline command. - Other actions point to suggested memory changes, reviewed memory, and memory
quality. - The actions are links to existing Hub View surfaces. They do not create a new
write path or change the memory model.
Clearer Agent Connection Paths
- The agent-context page now separates the main paths:
- chatbot copy/paste
- Codex one-click local setup
- Claude Code setup command
- Hermes/custom startup rule
- generic MCP config shape
- terminal fallback
- Codex is shown before Claude Code because Hub View can provide a guarded local
setup action when the project folder is known. - The terminal command is explicitly framed as a temporary fallback, not the
intended daily workflow. - Long commands and local paths wrap cleanly on desktop and mobile.
Connection Verification
- Hub View now shows a Connection verification block before the setup cards.
- The verification block is visually separated from the setup cards and uses
short status language so the difference between checked and external setups is
easier to scan. - Codex setup has a real local status:
Codex setup verifiedCodex setup not installed yetDemo preview onlyCannot verify yet
- Hub View can verify Codex only by inspecting the repo-local
AGENTS.mdblock
in a registered project folder. - Claude Code, Hermes/custom agents, generic MCP agents, and chatbots stay
honestly labelled as manual or external checks. ADH does not claim it can
inspect another agent's private configuration or prove that an unconnected
agent read the context.
Recommended First Run
From a fresh folder:
git clone https://github.com/AlexanderSmyslowski/central-agent-data-hub.git
cd central-agent-data-hub
scripts/first_run_demo.shThen open Hub View and use:
- Hand context to an agent
- Create context pack
- Connection verification
For a non-blocking check without starting Hub View:
scripts/first_run_demo.sh --no-hub-viewFor a phone preview on the same trusted Wi-Fi:
scripts/first_run_demo.sh --mobileVerification
This release was checked with:
git diff --check
bash -n scripts/*.sh
.venv/bin/python -m compileall agent_hub
.venv/bin/python -m pytest -q
scripts/smoke_public_demo.sh
.venv/bin/python -m agent_hub.cli status
.venv/bin/python -m agent_hub.cli checkThe final short first-run observation confirmed that a tester could find
Hand context to an agent, use Create context pack, recognize
Connection verification, and explain the difference between chatbot
copy/paste, Codex local setup, Claude setup, Hermes/custom startup rules,
generic MCP config, and the terminal fallback.
Boundaries
- no schema change
- no migration
- no new dependency
- no new Hub-memory write path
- no autonomous agent execution
- no hosted service
- no MCP write tools
- no authentication or role system
Release Position
v0.1.7 is suitable for evaluating Agent Data Hub as a local technical preview
where the question is not only "what reviewed memory exists?" but also "can I
see how that context is handed to an agent, and what setup state is actually
verified?"
It remains a technical preview. This release improves Hub View guidance and
connection visibility; it does not turn ADH into a hosted app or guarantee that
an unconfigured external agent used the context.