Design Space Map

Morphological analysis of 12 design dimensions — chosen path highlighted

Morphological Matrix
Design Configuration
Dependencies
Rejected
#DimensionABCDChosen
1IR as Compilation TargetDirect → Runtime→ flowJson → Runtime→ IR → Multi-Target→ IR → Single-TargetC Multi-Target
2Architecture LayersTwo LayersThree LayersFour LayersController OverlayB Three Layers
3Agent CommunicationSingle report_observationMultiple FunctionsFree-Text ParsingStructured OutputA Single Function
4Evidence Signal ModelDiscrete Kinds OnlyContinuous ConfidenceDiscrete + Confidence + ProcessRubric-Level JudgmentsC Full Taxonomy
5Policy ExpressionStructured DataCode (Python/TS)Prompt InstructionsRule Engine (DSL)A Structured Data
6State ManagementExplicit SM OnlyEvent Sourcing OnlyHybrid (SM + Events)Reactive StateC Hybrid
7Agent Autonomy GradientTwo LevelsThree LevelsFive LevelsFull AutonomyB Three Levels
8Event ProtocolPush OnlyPull OnlyEvent Sourcing OnlyHybrid Push + StoreD Hybrid
9Evidence OutputSeparate LedgerEmbedded in TranscriptPost-Hoc ExtractionHybrid Real-Time + EnrichmentA Separate Ledger
10Recovery StrategyFully AutomatedHuman-in-the-LoopCategorizedGraduatedC Categorized
11Transition AuthorityRuntime OnlyLLM Proposes + Runtime ApprovesLLM DecidesRuntime + LLM DelayB Propose + Approve
12Context ManagementRESET OnlyAPPEND OnlyHybrid RESET + SummaryPolicy-DrivenD Policy-Driven

The chosen configuration as connected design path:

1IR: Multi-Target
2Arch: Three Layers
3Comm: Single Fn
4Evidence: Full Taxonomy
5Policy: Data Objects
6State: Hybrid
7Autonomy: Three Levels
8Events: Hybrid
9Output: Ledger
10Recovery: Categorized
11Transition: Propose+Approve
12Context: Policy-Driven

Design Characteristics:

Determinism over Flexibility
Policies are structured data; transitions are runtime-approved; evidence is validated before recording
Separation of Concerns
IR ↔ Runtime ↔ Pipecat cleanly separated; evidence collection ↔ evaluation separated
Defense in Depth
Multiple constraint layers: IR policies, runtime guardrails, output validation, context policy
LLM as Sensor, Runtime as Judge
The LLM observes and reports; the Runtime validates and enforces
Auditability as First-Class
Events, evidence signals, and policy decisions all recorded with provenance

Hover a node to highlight its dependency chain. Hover an edge to see the dependency description. Amber nodes = hub dimensions (3+ connections).

Hover a card to see exactly which option was rejected (red strike-through) and which options are kept (blue). Each card shows a single-dimension slice of the morphological matrix.

Chosen option
Design characteristic
Rejected (fatal)
Hub dimension (3+ deps)