Architecture¶
This section describes the internal architecture of Lango, a Go-based AI agent built on Google ADK v1.0.0
-
High-level architecture layers, component interactions, and a visual diagram of how Lango processes messages from channels through the agent runtime to AI providers.
-
Complete directory layout with descriptions of every package, explaining what each module owns and how packages relate to each other.
-
End-to-end message flow, the bootstrap/wiring process, and async buffer patterns used for embedding generation and graph indexing.
-
Top-level product constitution, product path, capability areas, and execution-track portfolio for Lango.
-
The first audit ledger for the product area that most directly defines Lango: trust, pricing, settlement, teams, and shared artifacts.
-
Trust, Security & Policy Audit
Detailed audit ledger for the policy and safety boundaries that determine whether early knowledge exchange is actually safe to operate.
-
Identity Trust Reputation Audit
Audit ledger for identity continuity, trust entry, reputation, and revocation in
knowledge exchange v1. -
Pricing Negotiation Settlement Audit
Audit ledger for pricing surfaces, negotiation, settlement, and escrow in
knowledge exchange v1. -
The first transaction-oriented runtime control plane for
knowledge exchange v1, centered on transaction receipt and submission receipt with explicit current limits. -
:material-bank-check-outline: Settlement Progression
The first transaction-level settlement progression slice for
knowledge exchange v1, covering approve, revise, reject, and escalate with explicit current implementation limits. -
The first direct settlement execution slice for
knowledge exchange v1, connectingapproved-for-settlementstate to real payment execution with explicit current limits. -
The first direct partial settlement execution slice for
knowledge exchange v1, executing one canonical partial amount with explicit current limits. -
The first funded-escrow release slice for
knowledge exchange v1, connecting funded escrow and approved settlement state to real release execution with explicit current limits. -
The first funded-escrow refund slice for
knowledge exchange v1, connecting review-path funded escrow to refund execution with explicit current limits. -
The first dispute-linked escrow hold slice for
knowledge exchange v1, recording hold evidence for funded dispute-ready escrow with explicit current limits. -
Release vs Refund Adjudication
The first post-hold adjudication slice for
knowledge exchange v1, recording canonical release-vs-refund branching without yet executing either path. -
Adjudication-Aware Release/Refund Execution Gating
The first slice that connects canonical escrow adjudication to the existing release and refund executors with explicit branch gating and one-way safety rules.
-
Automatic Post-Adjudication Execution
The first inline convenience slice that lets escrow adjudication optionally execute the selected release or refund path immediately while preserving existing executor gates.
-
Background Post-Adjudication Execution
The first async convenience slice that lets escrow adjudication enqueue the selected release or refund path onto the existing background task substrate.
-
:material-refresh-alert: Retry / Dead-Letter Handling
The first retry slice for background post-adjudication execution, adding bounded exponential backoff and terminal dead-letter failure without changing canonical adjudication.
-
Operator Replay / Manual Retry
The first operator-facing replay slice for dead-lettered post-adjudication execution, reusing canonical adjudication and the existing background dispatch path.
-
The first authorization slice for replay, adding actor- and outcome-aware allowlists on top of the canonical replay gate.
-
Dead-Letter Browsing / Status Observation
The first read-only operator visibility slice for actor/time-aware dead-letter backlog browsing and per-transaction post-adjudication execution status.
-
The first concrete product track for external sovereign-agent economic activity.