ChatBI Studio v1.4.0 (formerly ChatBI V2)

ChatBI Studio v1.4.0

ChatBI Studio is the current user-facing product name of ChatBI V2. For v1.4.0 compatibility, the repository name, package/database/Compose/image identifiers, API and evaluation identifiers, and historical release names remain ChatBI-V2 / chatbi-v2 where changing them would break existing deployments.

ChatBI Studio v1.4.0 advances the governed ChatBI workflow from datasource connection and semantic modeling through verified natural-language analytics, charts, reusable answers, dashboards, and evaluation.

Highlights:

• Improved Windows one-click startup and bootstrap reliability.
• Added Backend-managed Excel/CSV datasource import and lifecycle controls.
• Aligned real SSE streaming with verified final-answer presentation.
• Improved semantic-model editing, differentiated dashboards, navigation, evaluation trends, model-service cards, and user/role controls.
• Hardened MiMo, DeepSeek, and Kimi routing while retaining server-side governance and verified fallbacks.
• Added public governance files, bilingual entry documentation, SBOMs, release/rollback documentation, sanitized screenshots, and CI coverage for main.

Validated locally with 817 passing Backend tests, 68 passing Frontend tests, 90 passing core browser tests, Golden 50/50, and dangerous SQL guard 56/56. Exact GitHub Actions URLs and final merged SHA are recorded only after publication.

This is an Apache-2.0 source release for local deployment, Enterprise PoC, private-deployment validation, and secondary development. It is not a production certification or a guarantee of external provider quotas, billing, or availability.

ChatBI V2 v1.3.1

ChatBI V2 v1.3.1

ChatBI V2 v1.3.1 is an Enterprise-oriented Open-source ChatBI / NL2SQL Platform source release. It supports local deployment, Enterprise PoC, private-deployment validation and secondary development. It is not a production-deployment certification.

Highlights

1. Fixed conversation intent routing, context isolation and stale-answer handling.
2. Improved chart/table rendering and automatic positioning for long answers.
3. Closed the MiMo (mimo-v2.5), DeepSeek (deepseek-v4-flash) and Kimi (kimi-k2.6) configuration, status and single-Model-Gateway path.
4. Added Query and Security Settings with explicit server-side enforcement boundaries.
5. Added Workspace, Appearance and read-only System Information views.
6. Added User, Role, Invitation, Audit and RBAC administration closure.
7. Preserved the independently certified actionable-control baseline: 894/894 PASS, with integration-delta controls PASS.
8. Added enterprise quick-deployment workflows for Doctor, Bootstrap, Start/Stop, Backup/Restore, Configuration and Troubleshooting.
9. Unified three supported modes: Default Open Source, Local Showcase and Enterprise PoC.
10. Advanced the unique Alembic migration head to 20260828_0013; the V1.3.0 rollback target remains 20260822_0012.
11. Passed Golden 50 execution, result-value and semantic/Oracle gates.
12. Passed two independent clean Enterprise deployment validations.

Certified release evidence

• Backend: 704 collected, 697 passed, 7 designed skips, 0 failed, 0 errors.
• Frontend: 64/64 tests, TypeScript diagnostics 0, production build PASS.
• Browser core experience: 45/45 PASS at the certified integration SHA.
• PostgreSQL and MySQL: connection, read-only enforcement, Schema/Catalog sync and representative query PASS.
• Exact-SHA live Provider smoke: MiMo, DeepSeek and Kimi PASS through the single Model Gateway with no direct-provider bypass.
• Migration: empty database to 0013, 0013 → 0012 → 0013, and existing 0012 → 0013 PASS.
• Fresh Enterprise deployment: 2/2 PASS.

The final Release SHA, annotated-tag object, GitHub Release ID/URL, checksums and exact publication timestamp are recorded in the external V1.3.1 Final Release Manifest because a Git commit cannot self-record its own immutable identity.

Known risks

• Graceful shutdown can still report exit code 137 for Backend/Sandbox Proxy while Docker reports OOMKilled=false. No corruption, resource leak or restart failure was observed; this remains registered and non-blocking for this source release.
• External Provider billing remains UNKNOWN_PARTIAL; the exact-SHA test ledger records the bounded calls and confirmed internal cost estimate, but does not claim complete provider-side billing reconciliation.

Production boundary

This release does not include production deployment certification. Kubernetes, Helm, HA PostgreSQL, multi-node disaster recovery, production monitoring/SLA, production key rotation, signed immutable production OCI, new Enterprise SSO development and Terraform are deferred. Additional enterprise hardening is required before production deployment.

Local Showcase credentials documented elsewhere are LOCAL_SHOWCASE_ONLY and must never be used as Enterprise defaults.

ChatBI V2 v1.3.0

ChatBI V2 v1.3.0

ChatBI V2 v1.3.0 is the formal open-source source release certified at commit 52db955fd67ebe592c289399a135528c13cb3e3d.

This release supports open-source source distribution. It is not a production deployment certification.

Highlights

Unified Model Control Plane

• Unified MiMo, DeepSeek, and Kimi routing through the Model Gateway.
• Shared cost accounting, retry, circuit breaker, fallback, trace, and SSE controls.

Semantic / NL2SQL

• Selected WrenAI and OpenChatBI reuse, with IBM execution evaluation.
• Schema/catalog synchronization, semantic models, projection contracts, SQL Guard, and Result Oracle verification.

RAG / Agent

• Owner-authorized legacy RAG reuse and selected DB-GPT runtime integration.
• Bounded multi-agent orchestration and allowlisted tool execution with traceability.

File / Multimodal

• CSV/XLSX, PDF/OCR, and image/vision workflows.
• Selected PandasAI runtime integration.

Product Governance

• Conversation, project, sharing, batch, feedback, and verified SQL capabilities.
• Cost, trace, model, evaluation, and governance dashboards.

Enterprise Hardening

• Workspace isolation, security gates, migration and rollback controls.
• Supply-chain audit, SBOM, dependency audit, and control certification.

Formal verification summary

• DATA100: 100/100
• Result value accuracy: 1.0
• MiMo NL2SQL: PASS
• DeepSeek NL2SQL: PASS
• Kimi required gate: PASS
• Real multimodal: 3/3
• Remote Phase 5: PASS — run 32998373760
• Remote Phase 4: PASS — run 32998381355
• Remote Phase 3 + IBM: PASS — run 32998387195
• Phase 6 final audit: PASS, blocker findings 0

Owner-approved delta

SQLBot direct upstream runtime reuse is not included in the formal runtime because of upstream license and packaging constraints. The release uses an approved clean-room capability path under the documented license exception. This must not be interpreted as “SQLBot direct reuse PASS.”

Production deployment disclaimer

• Open-source source release: SUPPORTED
• Production deployment: NOT CERTIFIED

Before production deployment, operators must complete production provider key rotation, production secrets/configuration, immutable OCI/wheel supply-chain closure, and the formal production-environment deployment gate.

Registered non-blocking risks

Major:

1. Production OCI/wheel immutable supply-chain closure remains open.
2. Performance raw-sample full recomputability remains open.

Minor:

1. The restricted proxy still depends on the Docker socket.
2. V1.3 tracked version-document identity cleanup is deferred as a post-release follow-up.

These risks do not block this open-source source release. No local Phase 6 evidence bundle or sensitive raw material is attached to this GitHub Release.