Repository navigation
Junjo AI Studio 0.85.0
Junjo AI Studio records and explains AI execution: model calls, tool activity, application state, and repeatable evaluation results. Studio 0.85.0 lets you follow one application Store across composed Junjo Agents and Workflows and see exactly which execution changed it.
Breaking upgrade: wipe application data
Use Junjo Python SDK 0.69.0 with this release. Telemetry contract 3 replaces contract 2; old telemetry is not converted.
Stop the old stack and wipe Studio application data, or select a new empty data directory. This includes SQLite databases, ingestion WAL, and Parquet telemetry. Users, API keys, developer tokens, evaluations, and execution history are discarded. Complete first-user setup, create new credentials, and update application clients.
Follow the tagged reset procedure. docker compose down --volumes does not clear the host-mounted application-data directory; Caddy certificate storage is separate.
What you can investigate
- Follow application state shared by an Agent, its direct Node tools, and conditional Workflow tools. Studio reconstructs each execution's own interval within the shared Store and attributes mutations to their actual writers.
- Inspect application state separately from the Agent's private runtime history. Isolated Subflows retain their own Store evidence.
- Navigate between Agent diagnostics, nested Workflows, raw traces, and state-transition writers without losing the selected execution or transition. Explicit null values remain visible.
- Query the same state and execution evidence through the UI and SDK/CLI. The OpenAI SDK example demonstrates model-call evidence alongside granular application state using OpenInference.
The minimal and VM/Caddy distributions pin Studio 0.85.0 with SDK 0.69.0. Their generated archives and mirrors are validated against the release source and published image digests.
Changes since Studio 0.84.1 · SDK release
Validated release provenance and artifact evidence
Validated Junjo AI Studio 0.85.0 release.
Canonical source: https://github.com/mdrideout/junjo@1b5a79882124f1b854eb17720013ac35476d25d9
Workflow: https://github.com/mdrideout/junjo/actions/runs/35537161169
Immutable image tags:
- backend:
mdrideout/junjo-ai-studio-backend@sha256:d1fbc4f01d88c207663cc597a51fdcb6d34c25248aa609f519b8ef9a874a35f6(tags0.85.0and1b5a79882124f1b854eb17720013ac35476d25d9) - frontend:
mdrideout/junjo-ai-studio-frontend@sha256:9841ec1f298e90cb35254d6171c652e7f2e0d78b48fb3a2c9a28f6b9d5d9c3c0(tags0.85.0and1b5a79882124f1b854eb17720013ac35476d25d9) - ingestion:
mdrideout/junjo-ai-studio-ingestion@sha256:8dc43db38f3af997295a3e7350ce5a3d82b2c94e6eb09f5613489cea8a61aa0a(tags0.85.0and1b5a79882124f1b854eb17720013ac35476d25d9)
Deployment distributions:
- minimal: archive
4bde9fbdde5448e278acf5e545c9b30b2a3abd626cdcda6b6e036ae87deed365, tree83c0544ed178e3823be3628a5db4d3d3fffd5083ed4e69a41c6556d8d55c9fa4, mirrormdrideout/junjo-ai-studio-minimal-build:master@3cf8dcd91a70a3b768896d81626d4c4d6632778b - vm-caddy: archive
5c922d872fd651d1bdde2344b2013ba38acb0207a595b3178f075960249fbbd3, tree835c0e7f773ea6bebe75bc8bb61cfa182d2b59bf78359e35964f61819ade4cd4, mirrormdrideout/junjo-ai-studio-deployment-example:master@09108901fd33ab866458b2f17f5a5e770866176f