Repository navigation
Releases: ordstrom-ai/tala
Releases · ordstrom-ai/tala
Release list
Tala 31.1.2 — Package cleanup and refreshed documentation
Documentation
- Introduce Tala as the Python library and developer toolkit for the TDM dialogue platform.
- Explain its capabilities, installation, and dialogue-service prerequisites.
- Move the detailed language-model example into a dedicated Requestor guide.
- Update the PyPI summary and canonical GitHub repository link.
Packaging
- Restrict the wheel to Tala runtime packages, excluding test packages, dialogue-domain examples, and scripts.
- Explicitly include required runtime schemas and templates, plus the xs3p stylesheet and its license.
- Remove the obsolete interaction-test extractor.
Maintenance
- Remove legacy Docker/Jenkins files, the Gerrit commit hook, and a tracked editor backup.
- Add CI checks for wheel contents, source-distribution tests, and installed-wheel runtime assets.
No changes to dialogue behavior or runtime dependencies.
Tala 31.1.1 — Ordström GitHub release workflow
This maintenance release completes the move of Tala’s source hosting
and publishing workflow to Ordström.
- Canonical repository: https://github.com/ordstrom-ai/tala
- Required GitHub Actions checks build, test and lint changes before merging.
- PyPI publishing uses Trusted Publishing instead of a stored API token.
- Publication verifies that the release commit passed the required main tests
and that the wheel and source distribution match the release version.
No application-code or dependency changes are included. The PyPI package
name remains tala, and its existing branding and project metadata are unchanged.
Release commit: 4d2401fda2ca638914fd90db3b33f5988f6b719d
31.1.0 — Structured GPT output support
Added
- Support for forwarding JSON Schema response formats through
GPTRequest
to Requestor. - A
completion_statusproperty exposing completion outcomes, including
completed, truncated, and refused responses. - Documentation and an example for the structured-output API.
Fixed
- Malformed structured JSON now raises
GPTStructuredOutputErrorinstead
of silently returning a default response or attempting generic JSON repair. - Response formats are included in request cache identity.
- Restored the missing 31.0.2 changelog entry.
Maintenance
- Updated development dependency pins for build, pytest, flake8, and pylint.
- Aligned repository guidance with the shared guarded Gerrit workflow.
Tala 31.0.2
Fixed
- Fixed XML compilation of proposition-specific
forgetand
forget_sharedplan items when their proposition child is surrounded
by whitespace or comments. Normally indented XML now compiles correctly;
compact formatting is no longer required.
Documentation
- Corrected the changelog’s release grouping and added missing entries
for earlier releases.
Tala 31.0.1 — DDD JSON schema fixes
Release notes
This patch release fixes the packaged DDD JSON schemas and adds validation tests to prevent schema regressions.
Fixed
- Removed leftover merge-conflict markers from the domain schema, preserving knowledge-precondition question (KPQ) support.
- Added schema definitions for in_* conditions, boolean operators, and supported condition proposition payloads.
- Fixed validation of predicate-only forget targets and goal-proposition targets.
- Corrected bundle schema references so human-readable DDD bundles validate with or without the ddd wrapper.
Testing - Added offline validation of all four JSON schemas, the mockup_travel fixtures, and DDD bundles.
- Added regression coverage for KPQ questions, conditions, and forget targets.
Tala 31.0.0 — JSON:API and DDD refactoring
Breaking changes
- Expand JSON:API serialization and deserialization for dialogue model
objects, including moves, plans, propositions, and iterator state. - Replace legacy DDD condition names with explicit
in_*conditions and
support nested boolean conditions. - Remove implicit service-action downdating. DDDs must specify explicit
downdate conditions where needed. - Remove the deprecated multi-command
talaCLI. Usetala-interact.
Improvements
- Replace regex-based semantic parsing with recursive-descent parsing,
improving support for nested expressions and error reporting. - Add multi-antecedent domain queries and structured query/iterator models.
- Harden domain orchestration validation and preserve AUTO TIS log format.
- Improve SSE connection timeouts, setup-failure handling, and thread teardown.
- Update the default GPT-4o deployment to
gpt-4o-2024-11-20. - Make MQTT tests independent of a live broker and improve test reliability.
Migration
Update DDD conditions and service-action downdate definitions to the new
explicit model. Use the corresponding TDM migration when adopting the
expanded JSON:API model support.
tala 30.0.1
- Fix: preserve service_action when parsing legacy SERVICE_RESULT propositions from TIS JSON (fallback to _service_action, _arguments, _ontology_name).
- Fix: parse legacy invoke_service_action payload variants (action, action_name, dict-shaped names) without losing action name.
- Tests: add parser unit tests for legacy service action fields and service result deserialization.
Tala 30.0.0 — Move Refactors, DDD Loading Improvements, and Cleanup
- Refactored move modules and ICM handling to simplify move creation and parsing.
- Updated DDD loading to support JSON bundles and named files reliably.
- Switched model accessors to properties for consistency across the codebase.
- Added allow_binding_yn_answers and KPQ support to DDD JSON/XML flows.
- Removed deprecated semantic logic from the SDK and cleaned legacy APIs.
- Improved tests around DDD loading, JSON parity, and interaction flows.
29.11.3
Release 29.11.2
Release notes (since 29.11.1)
- Hardened JSON:API parsing and tolerance for optional/omitted fields.
- Normalized JSON:API includes and moved versioning to meta with safer handling.
- Prevented JSON:API ID collisions and improved invalid link failures.
- Added JSON:API DDD snapshot coverage and removed legacy DDD JSON support.
- Added domain query modeling updates.
- Fixed bind YN question JSON round‑tripping to preserve boolean values.