The Problem
Mangrove conservation efforts in coastal Kenya lack centralized, accessible technical guidance for species identification, carbon accounting, and restoration planning. Conservationists, researchers, and local communities need a reliable reference for the five primary mangrove species and their ecological roles, but existing resources are fragmented across academic papers, government reports, and local knowledge bases without systematic integration.
What This Does
The Mikoko Guardian agent is a single-agent FastAPI application residing in mikokoguardian/agent.py with data initialization in mikokoguardian/init.py. It provides five core tools via the Google ADK framework: identifymangrovespecies, getsiteinformation, calculatecarbonstorage, planrestoration, and answergeneralquestion. The agent operates against a focused dataset of five Kenya coastal mangrove species (Rhizophora mucronata, Avicennia marina, Sonneratia alba, Ceriops tagal, Bruguiera gymnorrhiza) and five significant coastal regions. Carbon storage calculations integrate area and forest maturity estimates to produce equivalent carbon credit values. The architecture references Gemini 2.0 Flash for general knowledge integration, though the actual Gemini integration point is not evident in the two Python files present. A .env file is documented in the architecture diagram but does not appear in the root file listing provided.
How To Use It
Setup: This is a Python project using FastAPI. The source code consists of two files in mikokoguardian/. No pyproject.toml, requirements.txt, or setup.py is present in the root structure; dependencies would need to be inferred from the agent-venv/ virtual environment, which contains a Python 3.12 installation with standard data science and web packages. No build configuration files are visible.
Configuration: A .env file is referenced in the architecture diagram (mikokoguardian/.env) but is not included in the file structure provided. Environment variables for Google ADK authentication and Gemini API keys would be required but are not documented in the available files.
Running it: The entry point would typically be uvicorn mikokoguardian.agent:app assuming FastAPI conventions, but no server startup script or main.py is visible in the root. The agent type is documented as "Single Agent" with "Easy" complexity.
Real-World Use
A conservation NGO operating in Kwale County could use this agent to rapidly generate restoration plans for degraded sites. For example, invoking planrestoration with parameters for a 10-hectare site with moderate degradation would return a customized plan specifying Rhizophora mucronata as the primary species for shoreline stabilization, Avicennia marina for interior zones, estimated seedling requirements, resource timelines, and carbon credit projections. The calculatecarbon_storage tool could simultaneously quantify the site's mitigation potential in tCO2e, providing both ecological and financial justification for funding proposals.
Code Health & Issues
No test files detected - repository-wide untested code paths. No pytest, unittest, or similar test infrastructure is present. No CI/CD pipeline - no automated build/test gate in .github/ or visible CI configuration. No LICENSE file - unclear usage and redistribution rights from the repository root. No input validation visible - the agent tools accept parameters but no schema or pydantic models are evident in the two Python files, suggesting potential unbounded input paths. Secrets in version control risk - the documented .env file is absent from the file listing, but if present it could contain API keys depending on how the repository is managed. Dependency hygiene - no lockfile or explicit dependencies in the root; reliance on the agent-venv/ environment state without reproducible setup.
The Bottom Line
This is a focused, single-purpose agent for mangrove conservation education and planning in coastal Kenya. The codebase is minimal - two Python files and a documented architecture - which makes it quick to review and adapt, but the absence of tests, a license, and CI/CD means it is not production-ready without significant operational hardening. It would be well-suited for research prototyping, educational demonstrations, or as a baseline for a more comprehensive conservation platform, but should not be deployed in a production advisory capacity without validation, input sanitization, and proper licensing.
Who should use it: Environmental researchers, educators, and conservation NGOs prototyping mangrove-related AI tools who need a working reference implementation and are prepared to establish testing, licensing, and deployment pipelines independently.