Skip to content

Releases: prateekdevisingh/kakveda

Release Notes - v1.0.5

01 Mar 06:53

Choose a tag to compare

Release Notes - v1.0.5

Date: March 1, 2026

This note summarizes only documentation and positioning updates included for v1.0.5.

Scope

  • Refined Kakveda positioning as a unified open-source platform for:
    • infra monitoring,
    • observability (metrics/logs/traces),
    • AI/LLM/agent monitoring.
  • Added explicit narrative on the global gap Kakveda addresses:
    • failure memory,
    • pre-flight prevention signals,
    • unified telemetry workflow.

Documentation Updates

  • Updated README.md:
    • expanded "Latest Feature Rollup" timeline ,
    • consolidated pricing table for requested tools,
    • "What Makes Kakveda Unique" and easy setup guidance.
  • Updated docs/COMPARISON.md:
    • structured timeline of recent implemented scope,
    • expanded capability and pricing comparison matrix,
    • setup simplicity and pricing reference links.

Pricing/Comparison Coverage Added

  • Datadog
  • Splunk AppDynamics
  • Logz.io
  • LangSmith
  • Arize
  • Azure Monitor
  • MLflow
  • Kakveda + Netra (OSS baseline)

Notes

  • Pricing values are documented as public starting rates or published billing models and may vary by region, usage, and contract terms.

Release Notes v1.0.4

28 Feb 09:52

Choose a tag to compare

Release Notes v1.0.4

Date: 2026-02-28

Highlights

  • Added full Netra host observability coverage and stronger dashboard surfacing for infra + observability workflows.
  • Added standalone host-install documentation for kakveda-netra with clear integration steps for separate machines.
  • Added background runtime support for Netra so users can run without keeping a separate terminal open.

Added

  • Dedicated host install guide:
    • docs/netra-host-install.md
  • Dashboard observability enhancements:
    • diagnostics summary block
    • SLO + error-budget summary
    • inferred service map table
    • synthetic checks table
    • incident timeline table
    • basic forecast and correlation summaries
  • Metrics drill-down pages for observability and infra trends.
  • Netra background daemon controls:
    • --start
    • --stop
    • --bg-status

Changed

  • Improved chart rendering resilience (fallback rendering when external chart script is unavailable).
  • Improved card navigation to detail pages with full-card click behavior.
  • Improved threshold handling and per-agent persistence behavior in infra monitoring.
  • Updated documentation with:
    • “today vs tomorrow” capability narrative
    • expanded market comparison (Datadog, AppDynamics, Logz.io, and other suites)
    • explicit statement of current industry problem and Kakveda’s resolution.

Netra Notes

  • DASHBOARD_API_KEY remains mandatory for registration, heartbeat, and dashboard config synchronization.
  • Background process state files:
    • PID: ~/.local/state/kakveda-netra/netra.pid
    • LOG: ~/.local/state/kakveda-netra/netra.log

Version

  • Project version set to 1.0.4.
  • CLI version set to 1.0.4.

Release Notes v1.0.3

17 Feb 12:07

Choose a tag to compare

Date: 2026-02-17

Highlights

  • SDK-first integration guidance across examples and demo docs.
  • Root README updated with a minimal SDK integration recipe for any agent framework.
  • Legacy KakvedaGuard references reframed as reference-only.

Changed

  • Examples docs now prefer kakveda_sdk and KakvedaAgent for integration.
  • Legacy manual guides point to kakveda_sdk.guard.KakvedaGuard instead of removed helpers.
  • Updated warn URL examples to port 8105.
  • Added SDK quick integration section to the examples index.

Notes

  • Legacy docs remain for historical context; new integrations should use kakveda_sdk.

Release Notes v1.0.2

17 Feb 11:43

Choose a tag to compare

Date: 2026-02-17

Highlights

  • Introduced kakveda_sdk with KakvedaAgent as the unified integration surface.
  • Updated langchain agent demo to use the SDK (governance, events, dashboard registration).
  • Added Docker and runtime scaffolding for the demo.

Added

  • New SDK package: kakveda_sdk (agent.py, guard.py, registration.py).
  • Demo utilities and docs in examples/langchain-agent-demo/.
  • Demo Dockerfile and .dockerignore.

Changed

  • Demo apps now call KakvedaAgent.execute() and use model metadata.
  • Integration examples updated to show SDK usage.
  • Example docs reframed to emphasize SDK-first usage and mark legacy guard steps as reference-only.
  • Legacy manual integration guides now point to kakveda_sdk.guard.KakvedaGuard and avoid removed helpers.

Removed

  • Legacy demo helpers agent_registration.py and kakveda_integration.py (replaced by SDK).
  • Redundant event publisher helpers.

Notes

  • The legacy KakvedaGuard docs remain for reference; prefer kakveda_sdk for new work.

Kakveda v1.0.1 Release

09 Feb 10:46

Choose a tag to compare

Kakveda v1.0.1 Release Notes

Release Date: 9 February 2026

Overview

v1.0.1 is a patch release that fixes a startup crash on new laptops caused by a SQLite schema mismatch in the dashboard.

Fixes

  • Dashboard SQLite migration: Ensures scenarios and scenario_runs tables exist, and adds missing columns (scenario_id, scenario_code, expected_behavior, note).
  • Startup stability: Prevents sqlite3.OperationalError when /scenarios queries scenario fields on older databases.

What Changed

  • services/dashboard/db.py
    • Added best-effort migrations for scenarios and scenario_runs tables
    • Backfills missing columns for scenario metadata

Upgrade Notes

If you already have a running dashboard:

cd kakveda-v1.0
docker-compose restart dashboard

This triggers the migration on startup.

Breaking Changes

None.

Known Issues

None reported for this patch.

Kakveda v1.0.0

09 Feb 09:06

Choose a tag to compare

Kakveda v1.0.0 Release Notes

Release Date: 9 February 2026

🎉 Welcome to Kakveda v1.0.0!

This is the first production-ready release of the open-source Kakveda LLM Failure Intelligence Platform, marking a significant milestone in our mission to provide LLM systems with failure memory, pre-flight warnings, and system-level health visibility.


📋 Overview

v1.0.0 introduces:

  • 50 Comprehensive Test Scenarios for systematic validation
  • Enhanced Audit Logging with dedicated UI
  • Improved Agent Management with robust health checks
  • Better Ollama Integration with extended timeouts
  • Production-Ready Security with CSP headers
  • Comprehensive Documentation for failure intelligence concepts

✨ Major Features

1. 50 Test Scenarios Dashboard

A complete test suite for validating failure intelligence capabilities across all system components.

Scenario Categories:

  • Preflight Validation (Q1-Q10): Pattern detection, cold start behavior, threshold boundaries, false positive prevention, policy overrides
  • Failure Classification (Q11-Q20): Fingerprinting accuracy, multi-label detection, similarity scoring, version changes
  • Pattern Detection (Q21-Q30): Frequency tracking, temporal patterns, cross-app analysis, drift detection
  • Health Scoring (Q31-Q40): Composite metrics, trend analysis, recovery tracking, degradation alerts
  • Global Failure KB (Q41-Q50): Storage efficiency, retrieval performance, cleanup policies, versioning

New Features:

  • GET /api/scenarios-data: Returns all 50 scenarios with code, title, description, expected outcome, category, difficulty, and test prompt
  • Interactive dropdown UI for scenario selection
  • Real-time execution and validation
  • Detailed expected vs actual outcome comparison

2. Audit Log System

Comprehensive audit trail for all administrative and user actions with dedicated UI.

Features:

  • Complete action history (create, update, delete, admin operations)
  • User and timestamp tracking
  • Color-coded action types for quick scanning
  • Searchable and filterable logs
  • Persistent audit storage

New UI: /audit

  • View all audit events chronologically
  • Filter by action type, user, or timestamp
  • Admin-only access with role enforcement

3. Enhanced Agent Management

Improved agent health checking with automatic fallback and better error handling.

Improvements:

  • Dual Health Check Strategy: Try primary URL first, fall back to localhost if unreachable
  • Port Extraction: Automatically extract port from agent URL for fallback
  • Better Error Messages: Clear feedback on why health checks fail
  • CSP Header Updates: Allow connections to agent ports (8120, 8122)

Health Check Flow:

1. Try agent URL (e.g., http://docker-host:8120/health)
2. If fails, extract port and try http://localhost:{port}/health
3. Return clear success/failure message to dashboard

4. Improved Ollama Integration

Extended timeouts and better error logging for more reliable LLM interactions.

Changes:

  • Timeout Increased: 4s → 30s for /api/generate
  • Timeout Increased: 8s → 30s for generate_with_meta
  • Error Logging: Log Ollama failures before falling back to stub responses
  • Graceful Degradation: System remains functional even if Ollama is unreachable

Benefits:

  • Handles slow LLM responses without premature timeouts
  • Better debugging with explicit error logs
  • Seamless fallback for demo/testing scenarios

5. Security Enhancements

Updated Content Security Policy headers for production deployments.

CSP Changes:

  • Added connect-src allowlist: http://localhost:8120, http://localhost:8122, http://127.0.0.1:8120, http://127.0.0.1:8122
  • Enables dashboard to communicate with standalone agents
  • Maintains strict frame-ancestors 'none' for clickjacking protection

🔧 What Changed

Dashboard Enhancements

  • services/dashboard/app.py:

    • Added 50 scenarios data endpoint
    • Dual URL health check with localhost fallback
    • Extended Ollama timeouts (4s→30s, 8s→30s)
    • Better error logging for LLM failures
    • Updated CSP headers for agent communication
  • services/dashboard/templates/audit.html (new):

    • Comprehensive audit log viewer
    • Action type filtering with color-coded pills
    • Responsive table design
    • Admin-only access control
  • services/dashboard/templates/scenarios.html:

    • Enhanced scenario dropdown with 50 test cases
    • Category grouping for easier navigation
    • Difficulty indicators
  • services/dashboard/templates/agents.html:

    • Improved health check feedback messages
    • Better error display for unreachable agents

Agent Services

  • services/agent_echo/app.py:
    • Minor stability improvements

Data Files

  • data/failures.jsonl: Updated with test scenario results
  • data/health.jsonl: Current system health snapshots
  • data/patterns.jsonl: Detected failure patterns

📦 Installation & Upgrade

Fresh Installation

git clone https://github.com/yourusername/kakveda.git
cd kakveda-v1.0
docker-compose up -d --build

Access dashboard at http://localhost:8110

Upgrade from v0.2.3

cd kakveda-v1.0
git pull origin main
git checkout v1.0.0
docker-compose down
docker-compose up -d --build

No database migration required – fully backward compatible.


🧪 Testing

Run 50 Scenario Test Suite

  1. Navigate to Dashboard → Scenarios
  2. Select a scenario from the dropdown (Q1-Q50)
  3. Click "Run Scenario"
  4. Compare expected vs actual outcomes

Test Agent Health Checks

  1. Navigate to Dashboard → Agents
  2. Click "Test" button for any agent
  3. Verify health check succeeds (or displays clear error)

Validate Audit Logging

  1. Perform any admin action (create user, update project, etc.)
  2. Navigate to Dashboard → Audit (admin only)
  3. Verify action was logged with correct timestamp and user

🐛 Known Issues

None at release time.


🔄 Breaking Changes

None. Fully backward compatible with v0.2.3.


📝 API Endpoints

New in v1.0.0

  • GET /api/scenarios-data: Retrieve all 50 test scenarios with metadata

Existing Endpoints (unchanged)

  • GET /: Dashboard overview
  • GET /agents: Agent management
  • GET /scenarios: Scenario catalog
  • GET /runs: Execution history
  • GET /traces: Trace viewer
  • GET /audit: Audit logs (admin only)

🎯 Performance Metrics

  • 50 Scenarios: Complete test coverage for failure intelligence
  • Audit Trail: 100% action tracking for admin operations
  • Health Checks: Dual-strategy with automatic fallback
  • Ollama Timeout: 7.5x improvement (4s → 30s)

📚 Documentation


🤝 Contributing

See CONTRIBUTING.md for guidelines.


📄 License

Apache License 2.0. See LICENSE for details.


🙏 Acknowledgments

Thank you to all contributors and early adopters who helped shape this release!


🔮 What's Next (v1.1.0)

Planned features for the next release:

  • Webhook integrations for external failure notifications
  • Multi-tenant support with workspace isolation
  • Advanced pattern clustering with ML-based similarity
  • Export/import for GFKB portability
  • GraphQL API for complex queries

For questions or support, visit https://kakveda.com or open an issue on GitHub.

Kakveda v0.2.3 Release Notes

05 Feb 10:10

Choose a tag to compare

Kakveda v0.2.3 Release Notes

Release Date: 5 February 2026

Overview

v0.2.3 focuses on kids-agent integration stability and new-laptop onboarding reliability. All critical setup issues have been resolved with permanent fixes.

Key Features

✅ New Laptop Setup (No Problems!)

  • Dashboard now automatically repairs demo users on fresh setups
  • Admin login works reliably with admin@kakveda.local alias (avoids browser email validation blocks)
  • Standalone kids-agent integrates seamlessly with Kakveda stack without port conflicts
  • Comprehensive troubleshooting guide added to TROUBLESHOOTING.md

✅ Kids-Agent Integration

  • Network Fix: Standalone kids-agent now properly attaches to Kakveda docker network
  • Event Publishing: Traces now successfully flow from kids-agent → event-bus → dashboard
  • Port Conflict Resolution: Ollama port 11434 no longer causes binding failures
  • Service Registry: Agent health checks work reliably from dashboard

✅ Documentation

  • Updated README with kids-agent setup instructions
  • Added "new laptop" section with step-by-step integration guide
  • Clarified agent networking (docker networks vs host binding)
  • Enhanced TROUBLESHOOTING.md with login and agent setup recovery

What Changed

kakveda-v1.0

  • services/dashboard/app.py

    • Bootstrap "repair" logic ensures demo users exist and are active
    • Optional DASHBOARD_BOOTSTRAP_FORCE_PASSWORDS=1 env var for password reset
    • Added admin@kakveda.local alias alongside admin@local
  • services/dashboard/templates/login.html

    • Added novalidate to form to bypass browser email validation
  • docker-compose.yml

    • Kids-agent service now gated by profiles: ["kids"] (optional, non-blocking)
  • README.md & TROUBLESHOOTING.md

    • New "New Laptop Setup" section with verified step-by-step instructions
    • Recovery guide for admin login failures on reused DBs

kakveda-kids-agent (v0.1.1)

  • docker-compose.yml

    • Attaches to Kakveda's docker network (kakveda-v10_default external reference)
    • Ollama service renamed to kids-ollama (avoids name collision)
    • Removed host port binding for Ollama; uses internal network only
    • Environment variables use correct service hostnames (event-bus, dashboard)
    • Added extra_hosts mapping for host.docker.internal (Linux compatibility)
  • README.md

    • Integration section updated with correct API key generation steps
    • Event-bus publish endpoint corrected (POST /publish with {topic, event})
    • Note added about Ollama port conflicts and setup alongside Kakveda

Verified Scenarios

Fresh Laptop Setup

  • Kakveda + kids-agent both start without port conflicts
  • Admin login works with demo accounts
  • Kids-agent events appear in Kakveda dashboard within seconds

Standalone Kids-Agent

  • Runs independently, publishes to Kakveda stack on same host
  • Health checks pass from Agents page
  • Auto-registration with API key works reliably

Upgrade from v0.2.2

  • No database migration required
  • Existing agents and traces preserved
  • Demo users automatically repaired on startup

Breaking Changes

None. Fully backward compatible.

Known Limitations

  • Kids-agent Ollama model availability depends on host Ollama pull (or remote endpoint)
  • Dashboard event-bus subscriptions reset on container restart (in-memory, non-persistent)

Installation & Testing

Quick Start (New Laptop)

# 1. Start Kakveda
cd kakveda-v1.0
docker-compose up -d --build

# 2. Create API key in Dashboard → Projects → API Keys

# 3. Start kids-agent
cd ../kakveda-kids-agent
export DASHBOARD_API_KEY='<key>' AUTO_REGISTER=true
docker-compose up -d --build

# 4. Verify
curl http://localhost:8120/api/ask -H 'Content-Type: application/json' \
  -d '{"question":"Hello!","child_name":"Test","child_age":5}'
# Should see a response + trace in http://localhost:8110 Runs tab

Login (if blocked by browser)

  • Try: admin@kakveda.local / admin
  • Or force password reset: DASHBOARD_BOOTSTRAP_FORCE_PASSWORDS=1 on startup

Commit History

  • f85bd0f: Fix standalone kids-agent Docker network integration
  • 039f763: Add v0.2.2 release notes (kakveda-v1.0)

Contributors

  • Prateek Chaudhary (Kakveda + Kids-Agent)

Feedback & Issues

Report issues or feedback to project maintainers. Ensure to include:

  • Docker version and OS
  • Steps to reproduce
  • Docker logs: docker logs <container>
  • Database state (if login issue): presence of /data/dashboard.db

Status: ✅ Production Ready
Tested On: Linux (Ubuntu), Docker Compose v1.29.2
Next: v0.3.0 (remote agent registration, multi-project support)

This release hardens **first-time setup** and **new-machine onboarding**, especially when a user reuses an existing

05 Feb 08:16

Choose a tag to compare

Kakveda v0.2.2 — Release Notes (5 Feb 2026)

This release hardens first-time setup and new-machine onboarding, especially when a user reuses an existing data/dashboard.db.

Highlights

  • Demo login bootstrap repair

    • On startup, the dashboard now repairs demo users when the DB already exists (reactivates is_active, re-verifies is_verified, and ensures role mappings exist).
    • Adds an opt-in recovery switch to re-apply demo passwords when needed: DASHBOARD_BOOTSTRAP_FORCE_PASSWORDS=1.
  • Browser-friendly admin login

    • Keeps the promised OSS demo admin admin@local / admin123.
    • Adds a browser-valid alias admin@kakveda.local / admin123 for cases where browsers block admin@local (email format validation).
    • Login form now uses novalidate to avoid client-side blocking.
  • Agents are optional (no-break defaults)

    • Documentation clarifies that Kakveda runs fine without any external agents.
    • Recommends using Docker Compose profiles for optional agents to prevent compose failures when an agent folder isn’t present.

Docs

  • Adds troubleshooting guidance for “default admin login not working on a new setup” and recovery options.
  • Improves README guidance for adding agents without breaking default installs.

Key Files Changed

  • services/dashboard/app.py
  • services/dashboard/templates/login.html
  • README.md
  • TROUBLESHOOTING.md
  • docker-compose.yml

Upgrade Notes

  • If you reused an old data/dashboard.db and can’t login with demo credentials, set:

    export DASHBOARD_BOOTSTRAP_FORCE_PASSWORDS=1
    docker-compose up -d --build dashboard

    Then login and unset the variable.

  • Recommended admin login for browsers: admin@kakveda.local / admin123.

Full Changelog: v0.2.1...v0.2.2

Dashboard UX, Dataset Apps, and Warning/Health Fixes

05 Feb 06:12

Choose a tag to compare

Release v0.2.1 — Dashboard UX, Dataset Apps, and Warning/Health Fixes

Release date: 2026-02-05

Highlights

  • Datasets: app selector now reflects real app_ids (e.g. kids-app) instead of hardcoded demo apps.
  • Playground: clearer safety status (SAFE/WARNING/BLOCKED), shows "You asked" + "Answer", and links to filtered Warnings.
  • Warnings: analytics and filtering hardened (including block actions) for external agents.
  • Health: health scoring timeline wiring + chart rendering improvements.

What changed (selected)

  • Dashboard
    • Dynamic app discovery for selectors (runs/warnings/datasets).
    • Subscriptions/handlers for external events (e.g., persisted runs and child safety alerts).
    • Warnings page analytics improvements and app_id filtering.
    • Playground UI improvements (safety signal + prompt/answer display).
  • Health scoring
    • Health endpoints and JSONL persistence/volume wiring.

Operational notes

  • If you are running with Docker Compose, rebuild only the dashboard when updating UI/templates:
    • docker-compose rm -sf dashboard && docker-compose up -d --build --no-deps dashboard

Known limitations

  • Drift detection is not yet a dedicated, first-class capability in this repo (see docs/COMPARISON.md).

v0.2.0 - Agent Discovery & Integration

02 Feb 08:45

Choose a tag to compare

🚀 Kakveda v0.2.0 - Agent Discovery & Integration

✨ What's New

🔌 Agent Discovery & Registration

  • New API Endpoint: POST /api/agents/register - External agents can now self-register with Kakveda
  • Agent Registry: All registered agents stored in database with name, URL, description, and capabilities
  • Auto-Registration: Agents can register automatically on startup

📄 New Agents Page

  • /agents Route: New top-level page to view all registered agents
  • Agent Cards: Display agent name, status, capabilities, and base URL
  • Health Check Button: Check agent health directly from dashboard
  • Test in Playground: Quick link to test any agent in Playground

🎮 Playground → External Agent Integration

  • External Agent Dropdown: Select any registered agent in Playground
  • Direct Agent Calls: Playground can now call external agents via /api/ask
  • Dynamic UI: Model dropdown hides when external agent selected

📖 Complete Integration Guide

  • Step-by-Step Guide: How to connect any AI agent to Kakveda
  • Generic Format: Template with placeholders for Docker commands
  • Code Examples: Python code for sending traces and auto-registration
  • Environment Variables Reference: Complete table of required variables

🧭 Navigation Updates

  • Agents Link: Added to all dashboard pages in top navigation
  • Consistent Menu: Dashboard → Agents → Runs → Playground → Admin

📦 How to Upgrade

cd kakveda/kakveda-v1.0
git pull origin main
docker-compose build dashboard
docker-compose down && docker-compose up -d