// Project
Changelog
Release history and development roadmap. Uneven Agent follows semantic versioning.
The first public release is v0.4.0 — April 11, 2026. Versions v0.1–v0.3 represent the internal development history shipped as part of that release.
- Configless Workspace Architecture — Redesigned boot flow to run out-of-the-box with zero setup files, automatically optimizing configurations lazily.
- Multi-Agent Task Orchestration — Support for delegating complex tasks to isolated subagents with safe session boundaries and recovery limits.
- Flexible Reasoning Effort — Active support for o1/o3-mini reasoning parameter configurations across OpenAI, Google Gemini, and Anthropic Claude.
- Custom Personas & Skills — Dynamic configurations to configure workspace-level custom rules and personas dynamically appended to AI context instructions.
- Fast Path Cache — Real-time path caching database utilizing file-list-cache to enable instantaneous lookup updates and avoid redundant indexing scans.
- Unified Command Prefix — Standardized all binary entrypoints and commands under the unified uneven executable instead of uneven-ai.
- Remote Shell & Configuration APIs — Interception of slash commands in chat channels and config GET/POST endpoints for external IDE extensions.
- Fast Startup Scanning — walkbuilder scan optimization ignoring compiler artifacts and caching directories to reduce boot LOOKUP times to milliseconds.
- Prompt Optimization & Caching — Optimized context construction to leverage cloud-provider caching, significantly reducing token consumption and accelerating response times during interactive sessions.
- Smart Context Retrieval — Enabled the assistant to automatically locate and load additional files from the project as needed, ensuring it has all necessary context to solve complex tasks.
- Adaptive Session Memory — Automatically scales memory retention based on the capabilities of the selected model, preventing performance bottlenecks in local execution while maximizing depth on large cloud models.
- Streamlined Diagnostic Logs — Cleaned runtime logs and CLI diagnostic messages to ensure a clean, brand-neutral, and distraction-free workspace.
- Workspace Cleanliness — Automatically stores performance logs and reports in a dedicated directory, preventing clutter in the project root.
- Format & Parsing Fixes — Resolved stability issues where special characters in code files could cause parsing errors or layout breaks in the terminal.
- Local Fallback Backups — Implemented a native fallback backup mechanism under .uneven/obsolete to copy files before any askf destructive action (overwrite or delete).
- Flexible Recovery Paths — Preserves the original file extension and relative directory structure with unique timestamp suffixes, enabling simple manual restoration.
- Offline/Gitless Fail-Safe — Direct instructions for users to recover overwritten files manually when Git is not configured or automatic reversion fails.
- Engine Performance Optimization — Redesigned local workspace mapping and retrieval routines, slashing context preparation latency from ~200ms to less than 5ms (making CLI context preparation practically instantaneous).
- Workspace Mapping Speed — Optimized from ~120ms to ~4ms (30x faster).
- Vector Store Search (HNSW/Cosine) — Optimized from ~45ms to < 1ms (45x faster).
- Physical File Reading (RAG) — Optimized from ~15ms to < 0.2ms (75x faster).
- Lock-free Concurrent I/O — Thread-safe index access reducing concurrent I/O blocking from average/variable to zero.
- Stateful Data Analyst — Integrated sliding conversational memory into the Database Analyst, enabling continuous follow-up questions and query refinement in the same session.
- Token Optimization in Analysis — Applied smart context caching to data analysis schemas and rules, cutting token consumption by up to 90% on subsequent queries.
- Dedicated Remote Chat Route — Added a dedicated POST /chat endpoint to the Remote Shell, allowing IDE plugins to interact directly with the conversational assistant in-process.
- Self-Documenting API — Enabled a developer-friendly GET / route on the Remote Shell server that displays active routes, payload specifications, and authorization guidelines.
- Zero-Day Vulnerability Scanning — Implemented advanced file-level scanning to detect custom security vulnerabilities and coding threats natively.
- Context-Aware Efficiency — Consolidated intent classification and file-path routing into a single reasoning step, reducing API queries by 33% and cutting response times in half.
- Immediate Conversational Responses — Enabled instant, single-turn answers for general software questions, bypassing secondary context calls to save time and token usage.
- Smart History Hardening — Implemented in-place system rules and files map injection in local session memory, maximizing cloud prompt caching to save up to 76% on API billing.
- Optimized Local BI Analyst — Fully integrated local database configurations with automatic schema introspection, custom table gating, and read-only query approval constraints.
- System Resource Optimization — Redesigned background worker processes to run with optimized CPU utilization and safety thresholds.
- Visual Interface Redesign — Redesigned CLI welcome headers and diagnostic logs into a clean, modern, and borderless left-aligned terminal layout.
- CLI Command Standardization — Standardized all command parameters, activation utilities, and offline logs under the unified uneven executable.
- Background Process Daemonization — Resolved an initialization issue where the background proactive autofix service failed to start or detach properly under specific packaged distributions.
- Release Metadata Synchronization — Fully aligned internal version references, configuration fallbacks, and edge proxy distribution endpoints to ensure version compliance across all deployment targets.
- Global Execution Pathing — Resolved pathing and subcommand resolution errors (such as background indexing and remote shell sessions) when running global command installations in specific system environments.
- Background Process Resilience — Corrected a silent initialization failure that prevented proactive auto-fixing and directory watchdogs from starting correctly under background execution contexts.
- System Version Standardization — Synchronized internal release tags and build metadata to ensure reliable updater verification and environment checks.
- Context-Aware Dialogue — Restored deep conversational history in the interactive shell, enabling seamless follow-up questions and preventing memory drift.
- Dynamic Workspace Synchronization — Automatic cleanup of deleted files in the directory mapping to keep the agent's context strictly aligned with your local files.
- Clean File Generation Feedback — Replaced raw code stream printouts with elegant interactive loaders during file-writing tasks to preserve terminal readability.
- Exact Streaming Token Metrics — Accurate real-time token tracking and local consumption billing metrics across all AI execution layers.
- Specialized Framework Diagnostics — High-fidelity error parsing and automated coding strategies tailored for modern web application architectures and secure routing constraints.
- Enhanced Smart Contract Heuristics — Specialized diagnostics targeting gas-efficiency optimizations and standard safety rules on decentralized code files.
- Improved Prompt Lifecycle Management — Graceful input loop protection inside the Conversational Shell, guaranteeing stable TTY confirmations during automated file-creation sessions.
- Decoupled Explainer Engine — Optimized performance path for file-level reviews and summaries, slashing query latency and token overhead.
- Bring Your Own Key (BYOK) Pluggability: Seamlessly switch cloud providers (Google Gemini, OpenAI, Anthropic Claude) via standard configurations while retaining complete local ownership and control of API credentials
- Granular Resource Gating: Advanced schema access controls to restrict automated database actions to explicitly whitelisted tables only
- Native Telemetry Isolation: Absolute operational guarantee of zero source code tracking, zero telemetry transmission, and complete offline capability
- Native Delivery Architecture — Optimized distribution pathways with flat resource usage for direct execution.
- Global Edge Intelligence — Real-time licensing and high-speed distribution powered by dedicated edge networks for instantaneous deployment.
- Enterprise Memory Protection: Active memory pruning and strict resource isolation guaranteeing flat RAM consumption under continuous operation
- Cloud Intelligence Stabilization: Optimized multi-modal alignment with Gemini and premium external providers
- Token Economy Engine: Local-first orchestration reducing external cloud inference costs by up to 98%
- Business Intelligence & Dashboards: Natural-language data analysis with automated visual dashboard generation
- Context Intelligence: Agent file selection now resolves with significantly higher accuracy
- Reliable response handling: AI suggestions are extracted correctly across all scenarios
- Large project efficiency: Knowledge base now handles codebases with many files more efficiently
- Accurate path resolution: File references in agent responses resolve correctly
- Expanded platform support: Pre-built binaries now cover additional operating systems
- Attacker Attribution: Active pentest probes now silently transmit the origin IP for trace purposes
- Threat Detection: Terminal watcher identifies incoming Uneven probes and alerts the administrator
- License Hardening: Active Pentester mode now requires a Team tier license
- Shell Polish: Improved Portuguese intent classification and fixed false success log notifications
- 3-Layer Privacy Protection: Enhanced data safety suite with Schema Filtering, Semantic SQL Auditing, and Live Result Redaction
- Forensic Audit Logs: Detailed local history of AI actions and system fixes stored in structured Markdown format
- Autonomous Security Pentester: Validated engine for detecting hardcoded secrets, injections, and infrastructure header flaws
- Universal Data Context: Seamlessly index and query .csv and .xlsx files for automated data processing tasks
- Conversational Interface — launch an integrated natural language environment in any language directly on the free tier
- Hardware Optimization — automatic detection of accelerated hardware offloading capabilities with streamlined setup
- Expanded Capabilities — added commands for file explanation, automated documentation generation, and proactive task processing
- Uncapped Custom Inference — direct custom key compatibility providing complete usage flexibility
- Task Isolation — safe background job handling guaranteeing uninterrupted host system operation
- Stable Parallel Inference — robust parallel execution paths ensuring clean response delivery under load
- Configuration Resilience — absolute safeguard mechanisms preventing file modifications during system interruption
- State Preservation — continuous serial coordination protecting local knowledge base consistency
- Stable public release — production-ready across all core features
- Streaming AI responses — output appears token by token instead of waiting for the full answer
- Human-centric fix approval — alertOnly mode suggests fixes for your review before applying anything
- Dynamic autofix toggle — switch between autonomous and assistant mode mid-session with --autofix
- Faster knowledge base — significantly smaller index files and faster search queries
- SBOM generation (uneven sbom) — export a software bill of materials for your project
- Improved provider support — updated model defaults and correct API endpoints for all external providers
- Hardware Acceleration: Automatic GPU/CUDA offloading (32 layers max) for lightning-fast local inference
- Ethical Guardrails: Built-in defensive system that prevents malware, exploit, and cracking generation
- Advanced Document Intelligence: Native support for PDF, Word (.docx), Excel (.xlsx), and CSV Pro (industrial parsing)
- Strategic AI Recommendations: Preferred path for Local AI and Google Gemini Flash for privacy and cost-efficiency
- System Diagnostics: New `uneven info` command for real-time hardware, VRAM, and version auditing
- Memory Hardening: Silent model auto-unload (10min) and anti-loop safety (5-try ceiling)
- Local Data Analyst: Direct natural language queries over local CSV and Excel files without DB requirements
- Process Execution Safety — absolute coordination protocols preventing instance interference
- State Resilience — safe snapshot processing safeguarding project indexing during power interruptions
- Ordered Audit Trails — linear event sequencing ensuring flawless history consistency
- Resource-Aware Scanning — optimized hardware utilization preventing system overload during security audits
- Network Deadlines — strict execution timers guarding against hanging requests
- Index Cost Preview — real-time intelligence estimating resource allocation before committing
- Custom Table Gating — granular user control over analyzed database segments
- Spreadsheet Analytics — automated reading of direct Excel document folders
- Humanized CLI Assistance — plain-language instruction strings for all commands
- Pentester static mode (OWASP Top 10, secrets, CVEs, injections, headers)
- Pentester active mode with SHA-256 signed authorization scope
- Malware scanner: 8 categories, CI-compatible exit code (uneven scan)
- AI data analyst: natural language → SQL → Excel/dashboard (uneven analyze)
- CI/CD headless pipeline with GitHub Actions integration (uneven ci)
- Session state protocol (.uneven/session.json)
- Security report generator: HTML + Markdown output
- License system: free / Pro tiers with machine fingerprinting and offline grace period
- Proactive terminal watching & real-time monitoring
- Multi-language stack trace parser
- Retrieval-Augmented Fix (RAF) system
- Surgical code correction engine
- Autonomous file watcher with live re-indexing
- Advanced semantic retrieval engine
- Vector store integration with optimized HNSW search
- Deep indexing for projects and databases
- Native connectors for PostgreSQL, MySQL, SQLite, MongoDB
- Automated web documentation scraping
- Native core infrastructure stabilization
- High-performance bridge for local inference
- Initial CLI toolkit: uneven init, uneven ask
- Proprietary local brain support
- Multi-provider orchestration: OpenAI, Claude, Gemini, Ollama