Portae perspective
Work as cards and lanes
Requirements, architecture, tests, messages, defects, and execution artifacts stay readable for humans and operable for agents.
AI engineering workbench · one truth · many perspectives
Siloquy.com
One project truth, many working perspectives.
Siloquy.com is an all-inclusive agentic project-management and software-understanding workbench: requirements, architecture, code, dynamic UML, messages, execution, repository topology, defects, team context, and evidence stay in one inspectable model. Each surface is a perspective on the same truth, not another place for work to drift away.
One Library. One evidence chain. Many perspectives.
The dropdown is the product thesis.
Choosing a perspective changes the instrument, not the source of truth. The same Library can be read as requirements, code meaning, message traffic, execution flow, repository lineage, defects, team responsibility, or one unified workbench. The object remains the same; only the working view changes.
In the earlier language, the graph-oriented view carried the Siloquy name. In this version, Siloquy.com names the whole workbench; Meaning Graphs names the perspective where source code becomes dynamic UML, semantic graph, and evidence.
Library
Cards
Meaning Graphs
Messages
Execution
Evidence
Requirements to architecture
Portae
Requirements, architecture, tests, decisions, and implementation evidence become typed Cards with stable IDs and links.
Learn more →
Graphical code analysis and creation
Meaning Graphs
Code becomes navigable meaning: source evidence, Meaning Cells, dynamic UML, dependencies, flow, and generation context.
Learn more →
Decentralized communication
Messages
Human and agent conversations carry structured context, references, answers, and decisions beside the work they affect.
Learn more →
Agentic project management
Execution
Missions, pushes, assignments, completions, reviews, and agent work become visible flow with evidence attached.
Learn more →
Operational topology
Execution Graph
The same mission state becomes a graph of dependencies, blocked work, ready work, support cards, and closeout paths.
Learn more →
Repository truth
Repository Topology
Branches, commits, merges, release inclusion, and runtime identity become visible beside requirements and execution.
Learn more →
Quality and recovery
Defects
Failures, regressions, verification cases, repair assignments, and closeout evidence stay first-class parts of the model.
Learn more →
Responsibility map
Team
Humans and agents are visible operators with roles, current focus, availability, and accountable work surfaces.
Learn more →
One shell
Unified
The shared shell lets the user move between perspectives while preserving canonical source-of-truth boundaries.
Learn more →The same truth can look like a board, a graph, or a conversation.
The important promise is not a single perfect screen. The promise is that every screen keeps its references back to source, requirements, decisions, agent actions, repository state, and verification evidence.
Execution and Execution Graph are the cleanest example: the board shows what moves next; the graph shows how that movement sits inside the mission.
Meaning Graphs perspective
Code as navigable meaning
Large repositories can be read through source evidence, dynamic UML, dependencies, flow, and meaning cells.
Execution perspective
Work as an operating board
Missions, pushes, assignments, and completions make movement visible without separating it from requirements and evidence.
Execution Graph perspective
Work as mission topology
The same execution truth becomes a graph of gates, dependencies, linked cards, and closeout paths.
“Every summary should resolve to evidence: source, requirement, decision, agent action, and commit.”
Foundational paper
Read the Siloquy Introduction
Why AI-written code needs a semantic layer above implementation.
Technical explainer
From one nerd to fellow nerds
What Siloquy and Portae are for when large codebases, agent work, and traceability meet.
Business example
From business idea to working dashboard
How non-programmers can use Portae and named AI Officers to shape a KPI dashboard safely.
Unified shell · conversational operation
Ask in one place; reveal in the right perspective.
The assistant layer should not become a hidden second truth. It can guide a human through the GUI, analyze software, dispatch agents, prepare architecture, and move toward code creation through diagrams, while the durable result is written back to the canonical model.
- Turn intent into a visible view plan instead of forcing the user to choose the right surface first.
- Operate Portae, Meaning Graphs, Messages, and Execution without losing review context.
- Move from diagrams to architecture and implementation with source-backed evidence.
Show the work behind this defect and what it blocks.
I can reveal the defect, related requirements, execution assignments, commits, and discussion in the right perspectives.
Intent to view
Preview before mutation
Evidence after action
Where the work stands today
Work is ongoing, and not every feature described on this page is available yet. The stable commitment is the model: every new perspective must attach back to the same project truth, rather than becoming another isolated tool with its own memory and its own story.