v1.1.0

cascadeflow v1.1.0 — Hardening Release

Highlights

• 10x faster import: ~20ms via PEP 562 lazy loading (down from ~1900ms)
• Proxy security: CORS opt-in, 10MB body limit, Bearer auth with constant-time comparison
• Thread safety: Locking for SDK patch/unpatch, run counters, and record()
• Performance: Pre-compiled regex in ComplexityDetector (~4x faster)
• DX improvements: Actionable ImportError for missing integrations, removed deprecated CascadeAgent params, langchain optional extra

Breaking Changes

• CascadeAgent no longer accepts deprecated config, tiers, workflows, enable_caching, cache_size, enable_callbacks params
• Proxy CORS default changed from * to None (opt-in via cors_allow_origin)
• rich moved to optional extra (pip install cascadeflow[rich])

npm Packages

• @cascadeflow/core 1.1.0
• @cascadeflow/ml 1.1.0
• @cascadeflow/langchain 1.1.0
• @cascadeflow/vercel-ai 1.1.0
• @cascadeflow/paygentic 1.1.0
• @cascadeflow/n8n-nodes-cascadeflow 1.2.0

Full Changelog

v1.0.0...v1.1.0

v1.0.0

What's Changed

• feat(n8n): tool call validation, domain verifiers, agent domain routing by @saschabuehrle in #125
• fix(n8n): eager model resolution for visual flow highlighting by @saschabuehrle in #127
• fix(n8n): single getInputConnectionData call for correct model resolution by @saschabuehrle in #129
• fix(n8n): shorten domain labels, single domain verifier toggle by @saschabuehrle in #130
• fix(n8n): move domain verifiers toggle to visible position by @saschabuehrle in #131
• fix(n8n): section dividers, tool call validation default on by @saschabuehrle in #132
• fix(n8n): rename to Domain Cascading, clarify verifier fallback by @saschabuehrle in #133
• feat(n8n): convert CascadeFlowAgent to standalone execute() node by @saschabuehrle in #134
• feat(paygentic): ship opt-in integration with resilient delivery by @saschabuehrle in #136
• fix(vercel-ai): support AI SDK parts messages and improve integration DX by @saschabuehrle in #138
• Fix LangChain runtime issues and align JS integration with 1.x by @saschabuehrle in #135
• feat(core): TS rule-engine parity and OpenClaw decision-log opt-in by @saschabuehrle in #139
• fix(core): refresh Anthropic factory defaults to current model IDs by @saschabuehrle in #128
• chore(testing): extended E2E benchmark session runner by @saschabuehrle in #106
• LangChain integration: add CascadeAgent loops + domain policy routing by @saschabuehrle in #137
• docs(langchain): refresh Anthropic model IDs in examples and guides by @saschabuehrle in #140
• chore(vercel-ai): bump package version to 0.7.3 by @saschabuehrle in #141
• fix(vercel-ai): refresh Anthropic IDs and strengthen integration DX by @saschabuehrle in #142
• Improve Vercel/LangChain integration docs, publish hardening, and deployed /api/chat E2E by @saschabuehrle in #143
• fix(proxy): resolve stats timeout and normalize tool-call shape by @saschabuehrle in #144
• Improve Vercel + LangChain integration loops and DX by @saschabuehrle in #145
• Add ClawHub CascadeFlow skill under OpenClaw integrations by @saschabuehrle in #147
• Harden v1 security posture and safe example tooling by @saschabuehrle in #146
• Harden ClawHub CascadeFlow skill metadata and safety guidance by @saschabuehrle in #148
• Clarify ClawHub skill credential metadata and package verification by @saschabuehrle in #153
• Improve integration example DX and tool-loop reliability by @saschabuehrle in #154

Full Changelog: v0.7.1...v1.0.0

Release 0.7.1

• fix: add drafter+verifier Anthropic models to all presets (242866d)
• fix: use Sonnet 4.5 as verifier in anthropic-only config (9cc5f9d)
• Restore anthropic-only preset to Opus-first verifier defaults (3989b7c)
• bump all packages to v0.7.1 (#124) (97b234b)

Release 0.7.0

• docs: update README with 15 domains and add Banking77 benchmark (8e17e44)
• fix: correct version to 0.6.5 and update Banking77 benchmark (ca26798)
• fix: resolve Ruff lint errors in banking77_benchmark.py (223e777)
• docs: consistent cascadeflow naming + domain routing image (#94) (c6554a2)
• feat: Full Integration - Vercel AI SDK, n8n, Proxy, All Providers (#95) (a923702)
• fix: resolve proxy import conflict for CI (bd26bc1)
• fix: Black formatting + Python 3.9 compatibility (6bc7512)
• fix: correct AI SDK package names + regenerate pnpm-lock (0ac09c4)
• fix: black format errors.py (e8ae386)
• fix: ignore UP045 ruff rule for Python 3.9 compat (2ff1853)
• fix: resolve duplicate ValidationResult + ruff lint errors (450e2b3)
• fix: apply Codex CI fixes - clean imports, vitest config (530d4dc)
• feat(quality): v14 function call validation + agentic benchmark (#96) (6ce735e)
• fix: resolve core build and typecheck failures (#97) (24bb33d)
• OpenClaw integration: production readiness + OpenAI-compatible server hardening (#99) (da0a96e)
• feat(skills): add cascadeflow skill for OpenClaw/ClaWHub (404b0f3)
• fix(skills): make cascadeflow skill generic for ClaWHub users (be7b959)
• fix: add OpenAI-compatible streaming chunks (initial role + final finish_reason) (bf8627e)
• fix: OpenAI streaming compatibility + add working gpt-4o config (a1c2842)
• fix: GPT-5 streaming + OpenAI-compatible SSE format (494d7f5)
• fix(skills): update anthropic-only config - code/general/reasoning use Opus verifier (e1d1d10)
• feat: map heartbeat/cron to Haiku-only domains for cost efficiency (96d398d)
• fix(skills): sync anthropic-only config template (afacd10)
• feat(skills): add heartbeat/cron domains to all config templates (9eca566)
• fix(health): Return degraded status when no providers initialized (6b8e8be)
• fix: track routing decisions (cascade/direct) in streaming telemetry (118a689)
• feat(vercel-ai): useChat integration + agentic tool-loop parity (#101) (1344468)
• fix(benchmarks): repair MMLU runner + use GSM8K config (#103) (9a397f8)
• Fix OpenClaw GPT-5 Responses + Anthropic tool loop (#100) (1e8251a)
• LangChain v2: tool-risk verifier gating + tool-safe streaming (#105) (269c946)
• Gateway proxy consolidation + DX cleanup (#104) (21e30f1)
• feat(config): update anthropic-only preset with correct model names (e4c04e3)
• fix(config): use Haiku 4.5 as default drafter (6da94c4)
• fix(config): use Sonnet 4.5 for code/reasoning/creative drafters (f48c372)
• fix(anthropic): preserve non-empty text across multi-block responses (#107) (59a126b)
• feat(n8n): CascadeFlow Agent node + model highlighting (#102) (325e055)
• fix(openclaw): emit parser-safe streaming final chunk (#108) (329dd8f)
• fix(openclaw): use HTTP/1.1 protocol for OpenAI-compat server (6864fc2)
• fix(openclaw): add Connection: close header for SSE streaming (4261b11)
• fix(openclaw): include content in delta of final streaming chunk (f3a7b14)
• fix(streaming): discard draft chunks when cascading to verifier (7ba4898)
• test(openclaw): align final streaming delta assertion (#109) (adaabd8)
• fix(streaming): dont duplicate content in final delta chunk (24a8746)
• fix(openclaw): handle tool text streams and avoid mixed cascade output (#110) (7ae1b44)
• fix(openclaw): add stream_options.include_usage support for OpenClaw/pi-ai compatibility (#111) (44d2fb3)
• fix: OpenAI streaming spec compliance for all cascadeflow users (#112) (26c6087)
• fix: handle tool call events in streaming mode (#113) (b5be027)
• fix: surface upstream provider errors to OpenClaw users and normalize no-logprobs confidence (e1416cf)
• fix(alignment): add paraphrase floor to prevent 0% Anthropic acceptance (eab6fa4)
• fix(telemetry): track draft acceptance and cost in streaming mode (095f4d1)
• style: fix Black formatting and Ruff lint errors for CI (e0912bc)
• fix: record draft acceptance telemetry in stream_events (a90333f)
• Revert "fix: record draft acceptance telemetry in stream_events" (55c6832)
• fix(streaming): validate text drafts in tool pipeline instead of blind rejection (783a31e)
• feat(openclaw): decision trace JSONL + domain stats (#114) (3ffcffb)
• fix(telemetry): draft acceptance stats stuck at 0% for tool streaming (0933b91)
• style: fix Black formatting and ruff lint for CI (6ece809)
• feat(openclaw): native OpenClaw integration with domain routing and decision tracing (#116) (0ee7da7)
• Fix forced-verifier routing + benchmark cost accounting (#115) (5721893)
• chore: release cleanup — remove dev artifacts, fix docs and examples (#117) (351f616)
• fix: integration audit — 6 fixes across LiteLLM, n8n, LangChain, Vercel AI, OpenRouter (#118) (e2c8d80)
• fix(e2e): harden basic usage parity script for reproducible CI (#119) (c637291)
• feat(openclaw): add demo mode auth bypass with per-ip rate limiting (#120) (bb2d342)
• fix: release prep — npm publish pipeline, package metadata, and Python packaging (#121) (9f2cfe8)
• fix: update README — examples, docs, features, icons, provider count (#122) (2091a45)
• chore: release v0.7.0 — version bumps, release workflow fix, Vercel AI first publish (#123) (badc2af)

Release 0.6.5

• fix: add --no-scripts flag to n8n package publish (dd1154f)
• Migrate npm publishing to OIDC trusted publishing (#77) (2c01fe0)
• Fix documentation and examples for v0.7 (#78) (fbaa106)
• Fix TypeScript API naming in root README (2d348b2)
• Fix reasoning-models.ts example: Update model names and add error handling (#79) (804cb3c)
• Benchmark: Provider Comparison (OpenAI vs Anthropic) (#81) (06e31ec)
• Add HumanEval code generation benchmark (#82) (a51e75f)
• Add benchmark infrastructure for performance validation (#80) (1fe1543)
• Add GSM8K math reasoning benchmark (#83) (9720333)
• Add benchmark runner for executing all benchmarks (#84) (0c43c40)
• Add MT-Bench multi-turn conversation benchmark (9c0df22)
• Add comprehensive benchmarking guide documentation (9cf0a22)
• Add TruthfulQA factual accuracy benchmark (6c00feb)
• Add customer support benchmark for real-world ROI validation (9bfe51d)
• feat(langchain-python): Complete LangChain tool calling and structured output support (#89) (143fecb)
• fix(langchain-ts): Match TypeScript quality scoring thresholds to Python (#90) (35314a7)
• fix(langchain): enable PreRouter by default to match Python behavior (#91) (5489f51)
• fix(ci): correct GitHub Actions syntax for secret checks (95a8afa)
• fix(ci): remove problematic if conditions from test-examples workflow (94c0962)
• chore: remove temporary development files and update .gitignore (#92) (9ad9d36)
• feat: implement full 16 domain configs with 2025 models (#93) (19576f9)
• feat(python): add enterprise HTTP configuration support (25fd5fa)
• feat(typescript): add enterprise HTTP configuration support (7479e17)
• docs: add enterprise HTTP configuration guide (f92121b)
• chore: update .gitignore for benchmark artifacts (a797bc8)
• style: fix Black formatting in config.py (83c35f4)
• feat(n8n): v2 with 16-domain routing and circuit breaker (f117069)
• fix(n8n): fix lint error in domain routing description (8a4b9cf)
• feat(n8n): dynamic domain model inputs and v0.6.2 (da8fecd)
• feat(n8n): individual domain toggles for better UX v0.6.3 (9ac96eb)
• feat(n8n): shorter domain input labels v0.6.4 (914c046)
• fix(n8n): disable semantic validation by default v0.6.5 (e918621)
• fix(n8n): remove semantic validation to prevent OOM v0.6.4-1 (4696c54)
• chore(n8n): bump version to 0.6.5 (364ec0c)
• docs(n8n): update README with v0.6.x changelog, remove old warning (125f14c)
• chore(n8n): bump to v0.6.6 with updated docs (e9eb7b0)
• docs(n8n): add multi-domain cascading documentation (c4858e5)
• chore(n8n): bump to v0.6.7 with domain docs (c36d2d2)
• chore: release prep cleanup (c82b09d)
• fix: correct year to 2025 in ARCHITECTURE.md (cb8d7ae)
• docs: update presets.md version to v0.6.0 (54fe772)
• chore: update pnpm-lock.yaml for @cascadeflow/ml ^0.6.0 (6054e6d)
• refactor: reorganize benchmarks into dedicated directory (5d56b2b)
• refactor: move benchmarks to tests/benchmarks/ (c6d48ab)
• docs: add benchmark results to README header (5d3c5a4)
• docs: format benchmark results as table (b4b6106)
• docs: simplify benchmark line in README (197369d)
• docs: adjust benchmark wording (be827c8)
• docs: change GPT-4o to GPT-5 in benchmark header (007c1ad)
• docs: add spacing around benchmark line (4b90be5)
• docs: add br tags for more spacing around benchmark (305df2d)
• fix: remove extraneous f-prefix from print statements (169fcd9)
• chore: bump all packages to v0.6.5 (02fc8e5)

Release 0.6.0

• revert: restore original logo sizes (cae1720)
• fix: use absolute URLs for README logo to ensure GitHub rendering (db8ed61)
• feat: smart max_tokens defaults and n8n port mapping improvements (1c89dc8)
• fix: simplify logo to single img tag for better device compatibility (28bf4a3)
• revert: restore picture element for theme-aware logo display (0f28921)
• fix: use responsive logo with relative paths and percentage width (8dd6645)
• fix: apply responsive logo styling across all READMEs (eeba099)
• fix: swap n8n port mapping and improve model debugging (52f71c4)
• fix: show specific model names in all cascade flow logs (c1063a7)
• fix: simplify main README logo with fixed width styling (e6a1661)
• Revert "fix: simplify main README logo with fixed width styling" (f130b50)
• fix: update logo SVGs with embedded fonts as paths (b60be7a)
• feat: TypeScript Parity - Milestone 1 Complete (Foundation & Core Infrastructure) (#67) (0f4da85)
• feat(n8n): Upgrade n8n integration with advanced cascade features (#68) (7bb23e6)
• chore: remove package-lock.json and prevent npm lockfile (using pnpm) (0588dcf)
• chore: move LangChain integration plan to feature branch (9a41678)
• feat: LangChain integration with universal provider support and PreRouter (#74) (671be43)
• fix: update n8n boolean parameter descriptions to start with 'Whether' (f3f6a14)
• feat: add @cascadeflow/langchain-cascadeflow to publish workflow (7af6e8f)
• chore: exclude generated TypeScript API docs from git (75ff111)
• feat: add Python LangChain integration with cascade pattern (222307d)
• feat(langchain): move to integrations/ and add PreRouter + complexity detection (286bbdb)
• feat(langchain-python): add streaming support with optimistic drafter execution (18503b8)
• feat(langchain-python): add model discovery and helpers (b6264a8)
• feat(langchain): add GPT-5 cascade demo (7522270)
• fix: format Python code with Black (916e0c2)
• fix: apply Ruff linting fixes for Python LangChain integration (ba9b891)
• fix: apply Black formatting to langchain_model_discovery.py after Ruff changes (29fe978)
• fix: resolve mypy type checking errors in LangChain integration (03bf9e3)
• feat: LangChain Python integration fixes and multi-instance docs (#75) (f1fe9f6)
• fix: resolve Python code quality (Ruff) linting errors (11653cd)
• chore: bump version to 0.6.0 (8b10ba5)

Release 0.5.0

• feat: enable npm provenance and add @cascadeflow/ml to publish workflow (b744fd3)
• fix: update @cascadeflow/ml dependency and TestPyPI permissions (495e4ec)
• chore: update pnpm lockfile for @cascadeflow/ml@0.4.0 (deaf408)
• fix: build @cascadeflow/ml before core (dependency order) (2df439e)
• fix: skip prepublishOnly script for n8n package (already built) (6c7be62)
• fix: make publish workflow not fail on already-published packages (58da107)
• fix: replace pepy.tech downloads badge with shields.io PyPI badge (f53f2bb)
• feat: add PePy total downloads badge (24842b5)
• fix: remove deployment tracking from publish workflow (5538418)
• fix: restore environment tracking to show green deployments (454ccea)
• docs: fix broken documentation links and clean up outdated references (8526cc5)
• docs: remove outdated version references and future roadmap items (e8915d9)
• OpenRouter provider (#62) (4320b4d)
• fix: typescript types and linting (#63) (b036829)
• fix: resolve n8n package workspace dependency issue (v0.4.1) (#64) (7617993)
• Release v0.5.0: Multi-instance support, TypeScript parity, and bug fixes (780ebcb)

Release 0.4.0

• fix(core): update dependency from workspace to npm version (3cc2df7)
• ci: fix test dependencies (3473546)
• ci: fix test workflow (1423984)
• ci: fix ruff linting errors (c095c6e)
• ci: fix critical ruff linting errors (fd0036a)
• Update TypeScript examples for newer models (#53) (5df2339)
• ci: disable strict mypy checking temporarily (8715662)
• ci: disable additional mypy error codes (c2da6ea)
• ci: exclude integration tests from CI runs (ee6928f)
• style: format code with Black (cc95c6b)
• test: fix domain detection test assertions (79fc8b0)
• style: format test files with Black (8e71395)
• fix: remove trailing space from config.yml filename (Windows compatibility) (2eaf676)
• test: skip flaky timing test on Windows (8259bab)
• test: skip flaky timing tests on Windows and macOS (add311e)
• docs: update README badges (0a12f91)
• docs: reorder README badges - X and Stars at end (f3b219b)
• feat(n8n): migrate to scoped @cascadeflow/n8n-nodes-cascadeflow package (9a013b5)
• ci: update n8n package name to @cascadeflow/n8n-nodes-cascadeflow (19ad777)
• docs: update all references to use @cascadeflow/n8n-nodes-cascadeflow (e2994bd)
• docs: fix remaining n8n package reference in main README (a337ee2)
• chore: bump version to 0.4.0 (82c74b6)

Release 0.3.0

• chore: remove redundant .env.template file (7418c19)
• chore: bump version to 0.3.0 for public launch (cb6d206)

Release 0.2.1

• chore: update all URLs after repository rename to lowercase (235f043)
• chore: remove duplicate org README from cascadeflow repo (acdf26b)
• chore: remove development-only files and backups for public release (35a73f4)
• fix: add rich as required dependency (0ea4143)
• fix: include all subpackages in Python distribution (f3f0c0b)
• fix: add rich to requirements.txt to match pyproject.toml core dependencies (d568864)
• chore: update CODEOWNERS and CONTRIBUTING.md for strict PR workflow (#37) (271843e)
• test: verify branch protection blocks direct push (9f3cb90)
• Revert "test: verify branch protection blocks direct push" (#38) (7774066)
• chore: remove outdated .github checklist files from v0.1.1 (#39) (f8eadf4)
• chore: update CascadeFlow logo assets (#40) (537cf79)
• feat: resize logos and remove internal documentation (#41) (e12b5d0)
• fix: correct logo size to display 1/3 larger (#42) (052f09b)
• fix: correct logo display using width attribute instead of viewBox (#43) (9965758)
• docs: improve TypeScript and n8n README formatting and readability (9c14dcf)
• docs: standardize brand name to lowercase 'cascadeflow' across all files (e20e9f7)
• docs: audit and cleanup documentation structure (03f372a)
• feat: update to latest OpenAI pricing and improve model accessibility (6b71a96)
• docs: update quickstart guide to use GPT-4o-mini and GPT-4o (a82a3d4)
• docs: update providers guide to remove GPT-3.5 reference (da867fe)
• feat: upgrade Claude models from 3.5 to 4.5 series (a6bf14d)
• docs: clarify GPT-5 verification requirements (fa43478)
• feat: add proper GPT-5 support with max_completion_tokens parameter (84a7f67)
• feat: update basic_usage.py example to use GPT-5 verifier (549c36a)
• fix: complete GPT-5 support - reasoning model with proper parameter handling (71b8e2c)
• fix: update to latest OpenAI pricing (Jan 2025) (a3ce663)
• fix: calculate tokens from costs instead of accessing non-existent fields (c9a6e43)
• fix: increase timeout for reasoning models and optimize max_tokens (03e4e2e)
• revert: restore basic_usage.py to use GPT-4o instead of GPT-5 (1c8bc13)
• docs: update cost_tracking.md with latest OpenAI pricing (Jan 2025) (ffaab47)
• docs: update TypeScript ML section to match Python style (ce8b02b)
• fix: correct ML semantic validation API documentation (f0af961)
• feat: LiteLLM integration and cost validation with accurate README pricing (02fa6fa)
• docs: merge integration READMEs into comprehensive guide (#45) (18814cc)
• docs: add comprehensive README for enforcement examples (#46) (2cefb85)
• docs: improve examples README with complete coverage and navigation (#47) (7da88c3)
• docs: update TypeScript package READMEs (#49) (45f1888)
• docs: remove future/planned content from READMEs and update feature table (#50) (eda4ca1)
• docs: fix packages/ml/README.md API documentation (#51) (ac802a5)
• docs: fix broken references and formatting in documentation (#52) (5c8f4a0)
• fix: resolve TypeScript errors and update tests (#48) (5ec0fd3)
• fix: tool calling example node.js (a976a06)
• docs: comprehensive validation fixes for bullet-proof developer experience (9dc37d6)
• revert: restore examples/basic_usage.py to working version (f67af86)
• revert: restore examples/reasoning_models.py to working version (e321a8b)
• feat(typescript): add complexity detection and PreRouter logic (6fd160e)
• fix(typescript): correct cascade tracking in cost-tracking example (c895bbc)
• feat(typescript): add alignment scorer and LiteLLM integration (5cc5b3c)
• docs: add ML models documentation for GitHub Models tab (7cd073c)
• docs: update feature parity documentation and accuracy (0d9385d)
• chore: gitignore Python vs TypeScript comparison docs (4273db3)
• chore: gitignore release notes (RELEASE_NOTES_V0.2.0.md) (888ba8d)
• docs: fix Examples section links in docs/README.md (9778fe6)
• fix(typescript): correct cost calculation with longest-prefix matching (8b4a116)
• docs: update TypeScript examples README to match Python style (aba1e12)
• docs: remove 🆕 emoji from Reasoning Models table entry (#55) (8f942d2)
• docs: fix TypeScript validation examples and add TypeScript quickstart guide (3985895)
• docs: fix TypeScript quickstart examples in main README (8fb359e)
• fix(n8n): update node for compatibility with core changes (547f32b)
• chore: bump version to 0.2.1 for pre-launch validation (a3e0849)