#%@=   %   #   =+#   =   @       #%@=   %@=+#  @=+#%  =+#%@   #%@=
%   #   = #   =   @  +#  =       %   #  @        #      %    #    
@=+#     #    +#%@=  # @ +       @=+#   =+#%     %      @     @=+ 
=  %     %    #   +  %  +#       =   @  +        @      =        %
+   =    @    %   #  @   %       +#%@   #%@=+    =      +    =+#% 

DEV PORTFOLIO

01 // plattera

PLATTERA

Agent harness for deed transcription and mapping

Plattera is a fully agentic workflow for turning deed images into mapped, reviewable land outputs. Its custom harness runs AI agents across the full path: redundant image transcription, transcript editing and verification, source and evidence review, deed-to-IR representation, geometry compilation, map sanity checks, georeferencing, validation, and tiled-map rendering. Rather than asking a model for one answer, the system saves artifacts and provenance between stages, exposes live run events and human-feedback points, retrieves supporting material when needed, and surfaces missing dependencies or unresolved boundary questions. The user gets an auditable path from a deed image to a mapped result, including the work, evidence, and issues behind it.

PythonFastAPIReactLeaflet
02 // sinap

SINAP

Desktop text-to-speech and speech-to-text tool

A Tauri desktop app for using selected text and voice through system-wide hotkeys. It captures highlighted text and reads it aloud with OpenAI text-to-speech, and includes audio recording, voice activity detection, transcription, hotkey configuration, and a small playback interface.

TauriRustReactOpenAI TTS
03 // algent

ALGENT

Agent orchestration and graph workspace

Algent is a desktop control center and Python backend for running and reviewing agent workflows. It includes a cockpit for launches, events, timelines, and intervention; model and tool integration; specialized research and editorial agents; and a graph workspace that records typed operations, commits, and provenance for agent outputs.

PythonReactTauriLangGraph
04 // ohmega-monster

OHMEGA MONSTER

Automated, evidence-based AI newsroom

Ohmega Monster is the public publishing surface of an agent-run newsroom. Its workflows collect signals from feeds, search, and research sources; build research profiles and briefings; develop story angles; draft, review, revise, and package articles; produce charts and visuals when the evidence supports them; and publish approved work to the site and social channels such as X. Alongside full articles, it produces short radar posts, research briefings, and data-driven insight posts. Each published article includes a readable source record with the underlying sources, claim status, caveats, and evidence limits, making the research and editorial process inspectable rather than treating generated prose as sufficient proof.

Next.jsMarkdownEditorial pipeline
05 // ralph-engine

RALPH ENGINE

File-based agent-run engine

Ralph Engine runs worker and reviewer agent loops against projects that adopt a small directory contract. It creates isolated run branches, writes tasks, logs, state, and output artifacts to the project tree, and lets a person or control UI pause, stop, redirect, or resume a run through a control file.

PythonCLIPytestNDJSON
06 // prosada

PROSADA

Repository-native story planning tool

ProsAda is a desktop application and tooling system for planning long-form fiction in the story repository that owns it. It models books, acts, chapters, narrative units, threads, characters, places, and guidance; renders timeline and graph views; validates structure and layout; and keeps canonical data as version-control-friendly JSON and prose files.

Next.jsTauriFastAPIPython
07 // chesstera

CHESSTERA

Interactive chess position analysis

Chessterra imports PGN games into an interactive board and lets users step through positions or play moves. It calculates piece-level influence across the board and renders configurable heatmaps and overlays for White, Black, and net control.

Next.jsD3chess.jsZustand
08 // scriptara

SCRIPTARA

Multi-corpus text analysis dashboard

Scriptara is a desktop application for loading text files into named corpora and examining their vocabulary statistically. It produces frequency and normalized distributions, Zipf, log-log, and z-score views, and compares texts through shared-word intersections and pairwise overlap scores.

PythonPyQt5PandasMatplotlib