Releases: sawrus/code-obfuscator
Releases · sawrus/code-obfuscator
Release list
v0.5.0
[0.5.0] - 2026-03-19
Added
- New on-demand blackbox flow in Makefile:
make e2e-blackboxbuilds fresh MCP Docker image, reconfigures Codex MCP, runscodex exec, and validates the exactquery.pyoutput block. - New MCP/integration regression coverage for request-id mapping flow, strict unknown-field rejection (including
message_payload), and path-based SQL/Python identifier replacement.
Changed
- MCP API moved to server-side mapping by
options.request_id: client-side mapping inputs (manual_mapping,mapping_payload) were removed from active tool contracts. obfuscate_project/obfuscate_project_from_pathsnow persist request mapping in MCP memory;apply_llm_output/deobfuscate_*resolve mapping only byrequest_id.- Global obfuscation boundary logic now replaces mapped terms inside snake_case fragments (e.g.
bs_user_ids -> mmm_user_ids) while still avoiding replacements inside larger alphanumeric tokens. - README and test prompt/docs were updated to the request-id workflow and blackbox validation steps.
- Package version bumped to
0.5.0.
Stabilization
Language-aware detection of obfuscation terms for 10 languages: Python, JavaScript, TypeScript, Java, C#, C/C++, Go, Rust, SQL, Bash.
Pre-release
[0.2.0] - 2026-02-23
Added
- Language-aware detection of obfuscation terms for 10 languages: Python, JavaScript, TypeScript, Java, C#, C/C++, Go, Rust, SQL, Bash.
- Extended test fixtures in
test-projects/for all supported languages. - E2E multi-language roundtrip test with runtime validation via available compilers/interpreters.
- CI workflow for build/test/coverage via Makefile (
make ci). - Release workflow triggered by semver-style tags (
v*) with multi-OS binaries. - New
SAMPLES.mdwith practical business examples.
Changed
- Forward obfuscation term detection now considers file language and keyword filtering.
- Makefile enriched with
ciandrelease-artifactstargets. - SVT updated to stress mixed-language large trees.
Business Impact
- Enables safe code sharing with cloud/local LLMs (including Ollama) in polyglot teams.
- Reduces accidental leakage of product/domain identifiers across code, comments, and string literals.
- Preserves reversible mapping to integrate AI-generated changes back into production code.