Skip to content

Releases: GoogleCloudPlatform/cxas-scrapi

v1.8.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 31 Jul 21:05
f5d7444

1.8.0 (2026-07-30)

Features

  • cli: expose --single-bidi-stream flag to evals report CLI (e81878b)
  • cli: expose --single-bidi-stream flag to evals report CLI (93e8b7d)
  • evals: add JSON output format for combined eval reports (a857351)

Bug Fixes

  • address runtime NameErrors and callback eval failures in CI (32b5cf7)
  • copybara: address code review feedback (c49b6f0)
  • copybara: resolve pytype, testing, and struct issues for upstrea… (17aff2a)
  • copybara: resolve pytype, testing, and struct issues for upstream 1.7.0 sync (50379a4)
  • cuj-report: restore future import ordering in skill scripts (d6cf5a9)
  • Defer 1:1 agent push out of migrate.py to respect CXAS 100-agent cap (838c140)
  • deployments: restore two-version minimum for traffic splits (293337e)
  • evals: apply CL 953545678 fix for turn 0 initial utterance in simulation evals (7d60f9d)
  • evals: move EvalUtils static helpers back into class scope (7159ca9)
  • migration: dynamically strip unimported DFCX triggers while preserving standard Python decorators (4a1c209)
  • migration: dynamically strip unimported DFCX triggers while preserving standard Python decorators (53270bd)
  • migration: runtime NameError (DFCXFlowModel is not defined) when cxas CLI initializes (3297ac1)
  • migration: strip all DFCX trigger decorators and preserve original function names for Python tools (5c11aa8)
  • migration: strip all DFCX trigger decorators and preserve original function names for Python tools (510ddc1)
  • resolve trailing ruff lint checks and CI failures (266b882)
  • test: hermetic bidi session mocks & PyType signature compatibility (1b9f9da)
  • test: hermetic bidi session mocks & PyType signature compatibility (6d235a2)
  • tests: add missing return type annotation to test_session_id_extraction (bcc041d)
  • typing: manually upstream copybara type ignores and overrides (c737fcd)
  • typing: remove unused List import in traces.py (4841bc1)

v1.7.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 17 Jul 21:15
6e82369

1.7.0 (2026-07-17)

Features

  • linter: add S008 rule enforcing single llmPromptSecurity guardrail per app (ca377ae)
  • migration: convert DFCX test cases to CXAS TurnTestCase simulations (667e8cf)

Bug Fixes

  • deps: update junit-framework monorepo to v6.1.2 (bccb88e)
  • deps: update junit-framework monorepo to v6.1.2 (7d36d4b)
  • evals: prevent agent audio output file chopping in audio files (5e6a4c8)
  • evals: prevent agent audio output file chopping in audio simulations (ac6f18f)
  • migration: prevent browser launch in headless context during tests (4da8715)
  • migration: prevent browser launch in headless context during tests (d2b6fb9)
  • persist bidi websocket for audio simulations behind opt-in flag (6682a83)
  • persist bidi websocket for audio simulations behind opt-in flag (6fa0399)
  • setup: Fix skill setup issue (ed79706)

Documentation

  • agent-foundry: fix setup.sh comment to reference pyproject.toml (5c03cc7)
  • agent-foundry: fix setup.sh comment to reference pyproject.toml (f4f5caa)

v1.6.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 10 Jul 20:51
e6cdc61

1.6.0 (2026-07-10)

Features

  • core: add dynamic voice_config configuration for client-side TTS (244d629)
  • cuj-report: add ADK framework ingestor skill (0e62b1f)
  • cuj-report: add ADK framework ingestor skill (5d29f05)
  • cuj-report: add DFCX framework ingestor skill (1131552)
  • cuj-report: add DFCX framework ingestor skill (b625cac)
  • evals: add --capture-agent-audio flag to cxas evals report CLI (8dd9901)
  • evals: add --capture-agent-audio flag to cxas evals report CLI (22db38c)
  • evals: propagate voice_config parameter through simulate_conversation (08aaa27)
  • evals: Support Multilingual TTS for Simulations (75d9f85)
  • Expectations can be judged based on audio token (inline_data) (882e1a0)
  • insights: add CCAI Insights topic modeling, scorecard activation & deployment, analysis rules, and interactive HTML dashboard report (78a53ba)
  • migration: add interactive Agent xprs designer and utterance harvesting gated behind --experimental-agent-xprs (#325) (c875c40)
  • runner: support voice_config parameter in EnhancedSimRunner (d69fed6)
  • skill_eval: add 10 new evaluation scenarios, multi-phase latency profiling, and SUT environment isolation (999befb)
  • skill_eval: add generative scenarios and timeout/interactive fixes (3164fad)
  • skill_eval: add multi-phase latency profiling and timer telemetry (52f0034)
  • skill_eval: enforce strict SUT environment isolation with dynamic uv index credentials (2d8899e)
  • skill_eval: introduce automated Antigravity CES skill benchmark suite and scenarios (f0d4f6b)
  • skill_eval: introduce automated CES skill benchmark suite and scenarios (c911b18)
  • support deployments in scrapi-sim-runner (cb3916c)
  • support specifying the deployment in scrapi-sim-runner (c74feb6)

Bug Fixes

  • Add fake tools flag (e2582fd)
  • add optional flags to use tool fakes (1503fbe)
  • created expectations only flag to ignore goals and updated goal … (#322) (c58efa9)
  • deps: update junit-framework monorepo to v6.1.1 (5eae2a1)
  • deps: update junit-framework monorepo to v6.1.1 (75aba00)
  • evals: accept and forward progress_callback in reporting.py (a9513a6)
  • evals: accept and forward progress_callback in reporting.py (c735944)
  • evals: resolve linter errors and update test assertions for dynamic voice_config (260f8b2)
  • extract voice_config from test case in simulate_conversation (c3523ca)
  • extract voice_config from test case in simulate_conversation (33a0fc0)
  • improve simulation retry logic (0ed7785)
  • improve simulation retry logic (fbac7c7)
  • linter: resolve generic search description linter false positives and add type-safe JSON checks (ab98f17)
  • load common expectations (5b4eece)
  • load common expectations (c33b838)
  • persist bidi websocket for whole session (f2a63d4)
  • remove files not related to change (8d598f5)
  • reporting: add expectations_only to legacy run_all_evals wrapper (6d8dfe4)
  • reporting: add expectations_only to legacy run_all_evals wrapper (b399281)
  • revert goal prompt for simulations (8a99cac)
  • revert goal prompt for simulations (1deb629)
  • skill_eval: address PR #328 review feedback on Timer and add unit tests (41380de)
  • skill_eval: clean up SUT virtualenv on head close to prevent tmpfs inode exhaustion (9059a20)
  • tools: add fault tolerance around retrieve_tools in Tools.get_tools_map (81e6482)
  • tools: add fault tolerance around retrieve_tools in Tools.get_tools_map (272ef98)
  • tools: catch specific Google and requests exceptions instead of generic Exception (1e8ddcd)

Documentation

  • add IAM permissions guide and update navigation (5697ab5)
  • add user guide for A/B testing and traffic splitting with SCRAPI (2a55ad5)
  • document voice_config parameter for local audio simulations (25063e8)
  • Update linter rules documentation (e524788)
  • Update linter rules documentation (35bdd9d)

v1.5.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 26 Jun 21:58
5af0029

1.5.0 (2026-06-26)

Features

  • add voice background noise and randomized burst artifact injection (80258b8)
  • address PR review comments on noise injection (7b8e84e)
  • agents: add CUJ standardizer agent, refactor framework detector, and compile deliverables (49e5f55)
  • allow customizing models for running and for evaluating evals (eda8f24)
  • cli: add --overwrite flag to cxas pull (029edb4)
  • cli: add --overwrite flag to cxas pull (4d14e96)
  • cli: add --overwrite flag to cxas push (3d0287e)
  • cli: add --overwrite flag to cxas push (7b4b24e)
  • cli: add --timestamped flag to evals report (19e528d)
  • cli: Add no-input flag for headless execution (408a564)
  • cli: flag variable update attempts in llm-lint (e3bc802)
  • core: add WebSocket timeout protection, close diagnostics, and support for citations and custom payloads (aff41c8)
  • core: add WebSocket timeout protection, close diagnostics, and support for citations and custom payloads (ca29e81)
  • cxas-cuj-report-generator: enforce explicit time confirmation turns in voice prompts (f3c360e)
  • cxas-cuj-report-generator: enforce spelled-out verbal forms for times in voice responses (2f7ae2f)
  • cxas-cuj-report-generator: implement batch grading, prompt optimizer, and TESTING manual (f822537)
  • cxas-cuj-report-generator: integrate graph traversal rules, voice realism guidelines, and naturalness standards (3a0eecd)
  • cxas-cuj-report-generator: mandate root-level call fields, generic fallback greetings, and sign-offs (5853b19)
  • cxas-cuj-report-generator: mandate spelled-out forms for percentages, order IDs, and reward counts (b777c50)
  • cxas-cuj-report-generator: require polite voice markers on every agent turn (7ae2c77)
  • cxas-cuj-report-generator: standardize sign-off phrasing to Dining Service goodbye (3df284b)
  • deployments: add traffic split support (5806bba)
  • deployments: add traffic split support (ed10846)
  • evals: add ability to timestamp eval reports (e695201)
  • fix smaple names (2bf98d8)
  • guardrails: add design guide docs and local create scaffolding (d3ef0c4)
  • include global instructions and dynamic DIs to llm-lint (7ed70b3)
  • ingestor: enforce first-turn description validation in append_turn.py (12487f9)
  • lint: add I016 prose-state-machine rule (399d664)
  • lint: add V100-V104 variable-reference rules (#291) (8bfc92d)
  • linter: Add S007 linter rule to ensure sub agents only have one parent (b844aff)
  • linter: flag variable update attempts in llm-lint (c4bc3a1)
  • linter: Validate single-parent constraint for sub-agents (S007) (cfacdfb), closes #301
  • llm-lint: expand semantic linter with red-teaming & robustness scenarios (2b1b988)
  • llm-lint: expand semantic linter with red-teaming & robustness scenarios (9e7768a)
  • Make generated local templates pass lint by default (97ceabd)
  • make non-interactive and headless execution safe (#55) (c61f9ab)
  • make use_tool_fakes as a flag when running evaluations (3138fcb)
  • migration: always-on consolidation + grouping gate default (#231) (abec278)
  • migration: emit tabbed HTML analysis report on every migration (#207) (85bebc1)
  • migration: file-watch fallback for grouping gate (856248d)
  • migration: Grouping Review tab in analysis report (read-only) (3458903)
  • migration: Grouping Review tab in analysis report (read-only) (#227) (58c49da)
  • migration: HTML grouping confirmation gate (opt-in) (7f2692d)
  • migration: implement consolidated agents name-mapping throughout agent config (0c325c3)
  • migration: local-first Phase 8 — skip per-flow CXAS deploys until post-consolidation (05cf6a4)
  • move llm-lint from a skill to a package prompt and register as a cxas CLI subcommand (9f36128)
  • prompts: add rules for quotes vs back-ticks and variable formatting to llm-lint (8d00356)
  • report-generator: integrate export transcripts to JSON feature in HTML report UI (9e5b8ad)
  • supervisor: implement ingestor supervisor and dynamic batch execution daemon (827bc5e)
  • trace: add 'cxas trace search' for full-text conversation search (536fdb6), closes #236

Bug Fixes

Read more

v1.4.1

Choose a tag to compare

@kmaphoenix kmaphoenix released this 29 May 20:26
b5b37f3

1.4.1 (2026-05-29)

Bug Fixes

  • fix .gitignore to prevent packaging issues (952070e)

v1.4.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 29 May 16:08
0894288

1.4.0 (2026-05-29)

Features

  • add audio replay capability to combined evaluation reports (cf5b904)
  • add audio replay capability to combined evaluation reports (8ac99ac)
  • add cxas-loss-analysis skill for conversation loss and patterns analysis (b6f0a27)
  • Add llm-based linting to capture semantic errors. (6fab5a6)
  • Add llm-based linting to capture semantic errors. (34ab035)
  • add native support for CES_API_ENDPOINT and CES_TRANSPORT (0ecf7fa)
  • add native support for CES_API_ENDPOINT and CES_TRANSPORT (214f27a)
  • cli: add cli commands for list version and compare version (603612b)
  • cli: add cli commands for list versions and compare versions (ba92521)
  • cli: add GECX tools, callbacks, and variables list/delete subcommands (4ef9760)
  • improve loss analysis pipeline with time filters, server-side filtering, and performance optimizations (abecd37)
  • unify migration interfaces, implement migration profiles, fixes (#193) (0d57531)

Bug Fixes

  • app_name propagation in cxas create (d9b655a)
  • resolve lint errors and improve placeholder validation (ab5b009)
  • trace: correct GECX/CES Console URL query parameter route in console_url (9ace03c)

Documentation

  • add GoogleSearchTool schema and create tool schema generation w… (d9758cb)

v1.3.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 22 May 18:18
c21a5ad

1.3.0 (2026-05-22)

Features

  • add cxas trace for end-to-end conversation observability (a76c35c), closes #120
  • add DFCX conversation runner with live + history trace export (#128) (5fb8a71)
  • add optimize structure consolidator (21e9667)
  • add rich response payload support to slot filling framework (77edb75)
  • add widget tool linting and core SDK deserialization support (1863906)
  • added basic support for OpenAPI toolsets (7a8e8d0)
  • added fuzzy match for turn evals, case insensitive for contains turn evals (bd1969d)
  • bella_notte: event preemption, response propagation (570e6ff)
  • bella_notte: event preemption, response propagation, and expanded tool tests (a6631f8)
  • bella_notte: event preemption, response propagation, and expanded tool tests (1a2fc50)
  • bella_notte: per-agent DAG tools, multi-slot setters, 3-tier steer-back (f66f74b)
  • bella_notte: structured logging, config validation, then_direct… (d709a28)
  • bella_notte: structured logging, config validation, then_directive, engine hardening (903d187)
  • cli: Address PR 84 feedback and improve CLI robustness (552c9c9)
  • cli: Extend CLI functionality for iterating on deployments (160fab9)
  • Consolidate and improve the usability protocols, and refactor the scoping layer to be cleanly structured as a first-class agent. (4afb355)
  • Docs/slot filling guide (b4c44bb)
  • implement component linting (56a6be2)
  • implement strict local GECX linter rules and scaffolder guidelines (bc022c4)
  • implement strict local GECX linter rules and scaffolder guidelines (3d5fe4d)
  • promote migration helpers (#137) (4bbcd29)
  • SCRAPI CUJ Report Generator. (03b8bf4)
  • split DFCX→CXAS skill into resumable migrate/stage1/2/3 scripts (eeadbe6)
  • split DFCX→CXAS skill into resumable migrate/stage1/2/3 scripts (#130) (535299e)
  • support OpenAPI toolsets in SCRAPI (57c1058)
  • support session variable accumulation in audio modality and pattern file filtering (f9b452b)
  • support session variable accumulation in audio modality and pattern file filtering (539212c)
  • topology linker html preview (#138) (3c96ab5)
  • utils: introduce request bucket RateLimiter for API client and … (36b9f8f)
  • utils: introduce request bucket RateLimiter for API client and evals pacing (81a3ee2)

Bug Fixes

  • add widget tool linting and core SDK deserialization support (dc87441)
  • deprecation warning for datetime (12cc64d)
  • deps: update junit-framework monorepo to v5.14.4 (13617cd)
  • deps: update junit-framework monorepo to v5.14.4 (abb7ade)
  • deps: update junit-framework monorepo to v6 (2736e3e)
  • deps: update junit-framework monorepo to v6 (4906e01)
  • deps: update junit-framework monorepo to v6.1.0 (66e7567)
  • deps: update junit-framework monorepo to v6.1.0 (641f0d8)
  • evals/turn_evals.py and utils/gemini.py lint errors fixed (534c509)
  • implement thread-local Client in GeminiGenerate to ensure thread safety (d53f768)
  • linter: Quick lint fixes in the CUJ report generator skill. (f714910)
  • linter: validate rootAgent JSON file (#86) (fb36d51)
  • linting (c9129e8)
  • lockfile (67284d3)
  • make various fixes to optimize sim runner (e463141)
  • move migration modules to src instead of skills (251f0ac)
  • remove files with just import in skills (47ac0f7)
  • remove files with just import in skills (0523dbe)
  • reporting: handle multiline agent responses in html (38bb898)
  • resolve HTML link collisions and enhance turn metrics in combined report (92ced47)
  • resolve lint errors in app.py (cb3ac31)
  • resolve parallel simulation gRPC crashes and welcome audio disclaimer race (7edd307)
  • resolve ruff lint and format violations (bff83c2)
  • sort imports in append_turn.py (c1b5de4)
  • unit test fail, grouping review moved under migration, inquirerpy dependency added (cc85e09)
  • update lockfile and uv setup ([02a6aee](02...
Read more

v1.2.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 13 May 22:17
815f423

1.2.0 (2026-05-13)

Features

  • Add Slot Filling DAG Framework documentation and Bella Notte ex… (e8d9fa2)
  • Add Slot Filling DAG Framework documentation and Bella Notte example (b91d048)
  • bella-notte: replace with full CXAS app structure (055240b)
  • bella-notte: auto-confirm, stall gating, fresh-pending setter h… (8a021be)
  • bella-notte: auto-confirm, stall gating, fresh-pending setter hiding (b0ddd35)
  • cli: add directory support, async execution, runs flag, and modality support to evals report (819cb29)
  • cli: add directory support, async execution, runs flag, and modality support to evals report (36f07be)
  • cli: add directory support, filtering, and robust expectation check to evals report (9f39cfc)
  • cli: add directory support, filtering, and robust expectation check to evals report (6e247e9)
  • comprehensive DFCX to CXAS migration optimization, Stage 1 and 2, and minor enhancements (#93) (625f464)
  • Fix fresh_pending to detect new slots added mid-readback (3a5c17f)
  • implement GCS storage support for combined evaluation reports and add CLI flag (bf85a6c)
  • implement GCS storage support for combined evaluation reports and add CLI flag (c17b9a8)
  • integrate migration into Scrapi CLI (#59) (ee6320c)
  • update agent-foundry skill to use subagents to improve context management (a9a1227)

Bug Fixes

  • add missing google-cloud-storage dependency (7bef179)
  • Add missing google-cloud-storage dependency (12ba121)
  • add noqa comments for ruff lint (F821, I001) (747787d)
  • C002: shorten _get_args docstring to satisfy line length linter (a095f2e)
  • C002: use AST to count callback args instead of regex splitting (70d1e5d)
  • init-github-action nested app dirs (0710e2e)
  • linter: remove app.json tools check and update tests (a516fa6)
  • linter: remove app.json tools check and update tests (a3abc4e)

Documentation

  • add Design Guide, Patterns, and Tutorials sections (c836509)
  • agent-development: add team collaboration guide for multi-developer Git workflow (eabedd0)
  • agent-development: add team collaboration guide for multi-developer Git workflow (cd3c520), closes #90
  • Docs/design guide patterns tutorials (c40c375)
  • scope DAG slot manager state to prevent cross-DAG-agent contamination (4e5b0b0)
  • scope DAG slot manager state to prevent cross-DAG-agent contamination (7749ded)
  • skills: add multilingual agent patterns to cxas-agent-foundry s… (23b70e4)
  • skills: add multilingual agent patterns to cxas-agent-foundry skill (e827656)
  • skills: add speech rate and pacing guidance to cxas-agent-found… (4a85484)
  • skills: add speech rate and pacing guidance to cxas-agent-foundry skill (90835bc)

v1.1.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 05 May 22:04
5bdb4cd

1.1.0 (2026-05-05)

Features

  • Add --modality flag to cxas_sim_eval run_evals script (77bab91)
  • Add --modality flag to cxas_sim_eval run_evals script (4ea51ea)
  • add GCS export and unified reporting path (8bf2cd6)
  • add GCS export and unified reporting path (4670d77)
  • add progress bar for eval runs (3960546)
  • add progress bar for eval runs (f3d33d6)
  • Add run-session text CLI to start a test session for the app (ae13a05)
  • Add run-session text CLI to start a test session for the app (a6c36eb)
  • Add simulations to Golden conversion (61469b3)
  • Add simulations to Golden conversion (4645add)
  • Add support for providing environment.json file when branching app (3170a40)
  • complete refactor to Pydantic models, fix codebase porting bugs (#26) (a38648c)
  • create deployment updates (e7ba86c)
  • evals: capture and render agent transfers and custom payloads i… (f33b6b3)
  • evals: capture and render agent transfers and custom payloads in sim reports (46bf15e)
  • implement colab dashboard and CLI dashboard modules, and refactor visualization components (2d6f80b)
  • implement colab dashboard and CLI interface modules, and refactor visualization components (b0bd6f5)
  • improve colab migration logs (#44) (ff43d78)
  • linter: add I014 rule to warn when current_date is missing from instructions (68d6a3d)
  • linter: add I014 rule to warn when current_date is missing from… (97a7e53)
  • more verbose implementation for channel types and deployments (db79776)
  • rich progress bar. remove aliev-progress (737c84a)
  • rich progress bar. remove aliev-progress (5a61d7c)
  • support events via audio (d5adc0a)
  • support events via audio (f016606)

Bug Fixes

  • (docs) changed underscores in terminal command params to hyphens (5184d72)
  • add missing alive-progress (Scrapi-wide fix) and nest_asyncio (migration specific) dependencies to setup.py (57613b0)
  • add missing alive-progress dependency to setup.py (14357e1)
  • add missing voice config to app.json template in foundry skill (3821b09)
  • add missing voice config to app.json template in foundry skill (4b7d1a6)
  • add rich progress bar, remove alive-progress (7bc1113)
  • correct indentation for justification strings in TurnEvals class (9bfb352)
  • Fix bugs in init-github-action (a94c019)
  • Fix bugs in init-github-action (3d498bc)
  • ignore localhost in readme link check (699eb66)
  • linter: accept display names in S004 child agent references (0ca0727)
  • linter: accept display names in S004 child agent references (2313da9), closes #16
  • linter: C009 false-positives on dict[str, Any] annotations (cd40176)
  • linting (50dc27e)
  • linting (7983803)
  • lint: update ruff version and fix import sorting (c006eb0)
  • parse callback signatures via ast in C009 to handle dict[str, Any] (6534cbe), closes #56
  • rename app_id to app_name; remove unused code (59804b8)
  • resolve deployment crashes, finalize e2e migration parity with colab tool (with enhancements) (#37) (6a3ae12)
  • revert app_id for create_app method (bc41156)
  • scope TOOL_INPUT/OUTPUT justification to failure and update README (54ce6cd)
  • typing, imports and removed deprecated version_id (fd83967)
  • update app_id to app_name in CLI (aa65fdb)
  • update app_id to app_name in CLI (057f853)
  • update to use proper typing (c895cd8)
  • use regex to parse for events (e3c4d68)

Documentation

  • add Mercedes F1 fan agent PRD example (42a7d64)
  • add Mercedes F1 fan agent PRD example (d35b056)
  • fix mkdocs build warnings (b178ec2)
  • fix setup instructions across documentation (4bd9cce)
  • fix setup instructions across documentation (1b2e47c)

v1.0.0

Choose a tag to compare

@kmaphoenix kmaphoenix released this 22 Apr 05:44
41d0890

1.0.0 (2026-04-22)

Features

  • Init release of CXAS SCRAPI (5f37dea)

Miscellaneous Chores