Quartz 4

vm-osil-systems-catalog

4 min read

OSIL Systems Catalog — Visual Map

Three Mermaid diagrams: (1) categories+systems graph, (2) stack-layer applicability, (3) optimized integration timeline.

Diagram 1 — All 22 categories with their named production systems

graph LR
    subgraph CAT_RAG["1 · RAG Retrieval"]
        Ragas[Ragas]
        LlamaIndex[LlamaIndex ✓]
        Voyage[Voyage AI ✓]
        ColBERT[ColBERT]
        Cognee[Cognee]
    end

    subgraph CAT_FEW["2 · Few-Shot Evolution"]
        BootstrapFS[DSPy BootstrapFewShot]
        AutoCoT[AutoCoT pattern]
    end

    subgraph CAT_TOOL["3 · Tool Selection"]
        Toolformer[Toolformer]
        Gorilla[Gorilla]
        Bandit[Thompson/UCB Bandit]
    end

    subgraph CAT_COST["4 · Cost-Aware Routing"]
        RouteLLM[RouteLLM]
        FrugalGPT[FrugalGPT]
        PortkeyR[Portkey ✓]
        OpenRouter[OpenRouter ⚠ stale]
    end

    subgraph CAT_CONST["5 · Constitutional AI"]
        AnthropicCAI[Anthropic CAI ✓ native]
        Lakera[Lakera Guard]
        NeMo[NeMo Guardrails]
        GuardrailsAI[Guardrails AI]
    end

    subgraph CAT_DEBATE["6 · Multi-Agent Debate"]
        AutoGen[Microsoft AutoGen]
        MetaGPT[MetaGPT]
        CrewAI[CrewAI ✓ KB]
        LangGraph[LangGraph]
        Magentic[Magentic-One]
        MoA[Mixture of Agents]
    end

    subgraph CAT_ADV["7 · Adversarial Self-Play"]
        evolveRL[evolveRL]
        SPIRAL[SPIRAL]
        AbsoluteZero[Absolute Zero]
    end

    subgraph CAT_CURR["8 · Curriculum Learning"]
        AgentBench[AgentBench]
        WebRL[WebRL]
    end

    subgraph CAT_SC["9 · Self-Consistency"]
        CoTSC[CoT Self-Consistency]
        ToTVote[Tree-of-Thoughts voting]
    end

    subgraph CAT_SYNTH["10 · Synthetic Data Distillation"]
        SelfInstruct[Self-Instruct]
        EvolInstruct[Evol-Instruct]
        Genstruct[Genstruct 7B]
        Distilabel[Distilabel/Argilla ✓]
    end

    subgraph CAT_SCHEMA["11 · Schema Correction"]
        instructor[instructor jxnl]
        Outlines[Outlines]
        Guidance[Guidance MS]
        PydanticAI[Pydantic AI]
    end

    subgraph CAT_LH["12 · Long-Horizon Credit"]
        ReAct[ReAct foundation]
        MCTS[MCTS for agents]
        PRM[Process Reward Models]
    end

    subgraph CAT_EVAL["13 · Eval Infra ⭐"]
        Langfuse[Langfuse RECOMMENDED]
        Phoenix[Phoenix Arize]
        DeepEval[DeepEval]
        Promptfoo[Promptfoo]
        TruLens[TruLens]
        Patronus[Patronus]
        Galileo[Galileo]
        HoneyHive[HoneyHive]
    end

    subgraph CAT_CL["14 · Continual Learning"]
        EWC[EWC]
        LoRA[LoRA adapters]
        PLDA[PLDA]
    end

    subgraph CAT_FED["15 · Federated"]
        Flower[Flower]
        FedML[FedML]
        SupabaseShare[Supabase reflection pool]
    end

    subgraph CAT_PLAN["16 · Plan Decomposition"]
        LangGraph2[LangGraph]
        ControlFlow[ControlFlow Prefect]
        Burr[Burr DAGWorks]
    end

    subgraph CAT_AS["17 · Active Sampling"]
        modAL[modAL]
        ArgillaAS[Argilla ✓]
        DeepEvalAS[DeepEval AS]
    end

    subgraph CAT_OUT["18 · Conversation-Outcome ⭐⭐"]
        HubspotDS[HubSpot deal stage]
        SalesforceE[Salesforce Einstein]
        GongRI[Gong Revenue Intel]
        IntercomFin[Intercom Fin]
    end

    subgraph CAT_JUDGE["19 · LLM-as-Judge Tuning"]
        GEval[G-Eval]
        AlpacaEval[AlpacaEval]
        MTBench[MT-Bench]
        JudgeLM[JudgeLM]
        Atla[Atla judges]
    end

    subgraph CAT_CONV["20 · Production Conv AI"]
        IntercomFin2[Intercom Fin]
        Hippocratic[Hippocratic AI]
        Decagon[Decagon]
        Sierra[Sierra AI]
        Cresta[Cresta]
        Gong[Gong]
        Twilio[Twilio ✓]
    end

    subgraph CAT_VOICE["21 · Voice ⭐"]
        LiveKit[LiveKit Agents OSS]
        Pipecat[Pipecat]
        OpenAIRT[OpenAI Realtime]
        Vapi[Vapi]
        Retell[Retell AI]
        Bland[Bland AI]
        ElevenLabs[ElevenLabs Conv]
        Silero[Silero VAD]
        AssemblyAI[AssemblyAI]
    end

    subgraph CAT_HALO["22 · HALO Harness Opt"]
        HALO[HALO context-labs]
    end

    style CAT_EVAL fill:#fff3e0,stroke:#ff6f00,stroke-width:3px
    style CAT_OUT fill:#ffebee,stroke:#c62828,stroke-width:3px
    style CAT_VOICE fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px
    style CAT_HALO fill:#e3f2fd,stroke:#1565c0,stroke-width:2px

Diagram 2 — Stack-Layer Applicability

Where does each category apply across YOUR stack?

graph TB
    subgraph L1["Layer 1 · Claude Code main"]
        L1A[peterskoett SI skill ✓]
        L1B[Trajectory capture]
        L1C[AGENTS.md/CLAUDE.md]
        L1D[Eval infra Langfuse]
    end

    subgraph L2["Layer 2 · OpenClaw 36 agents"]
        L2A[GEPA prompt opt]
        L2B[Reflexion failure loops]
        L2C[Conversation-outcome]
        L2D[Constitutional AI for solara]
        L2E[Voyager skill induction]
    end

    subgraph L3["Layer 3 · 40 Skills"]
        L3A[Voyager auto-create]
        L3B[Promptfoo A/B testing]
        L3C[AgentSkills.io standard]
        L3D[Skill composition learning]
    end

    subgraph L4["Layer 4 · Infrastructure"]
        L4A[Cost-aware routing Portkey]
        L4B[HALO harness opt]
        L4C[Tool selection MCP routing]
        L4D[Schema correction webhooks]
    end

    subgraph L5["Layer 5 · Cron 70+"]
        L5A[Schedule optimization]
        L5B[Failure-mode learning]
        L5C[G-FAILED-SERVICE-MTTR auto-fix]
    end

    subgraph L6["Layer 6 · Future Voice Agent"]
        L6A[Voice infra LiveKit]
        L6B[Turn-taking opt]
        L6C[Sentiment-driven response]
        L6D[Latency-aware caching]
        L6E[HALO dialogue state machine]
    end

    L1 -.feeds.-> L2
    L2 -.uses.-> L3
    L3 -.routes via.-> L4
    L4 -.scheduled by.-> L5
    L5 -.invokes.-> L2
    L2 -.future.-> L6

Diagram 3 — Optimized Integration Timeline (7 waves)

gantt
    title OSIL Integration Timeline (56 mechanisms across 7 waves)
    dateFormat YYYY-MM-DD
    axisFormat %b %d

    section Wave 1 Foundation
    T1 capture (DONE)        :done, 2026-05-03, 1d
    T7 Cost routing          :2026-05-04, 7d
    T10 Eval infra Langfuse  :2026-05-04, 14d
    T38 Privacy Presidio     :crit, 2026-05-04, 10d
    T1.3 Trajectory capture  :2026-05-08, 7d

    section Wave 2 Headlines
    T3.1+3.2 DSPy+GEPA       :crit, 2026-05-15, 21d
    T11 RAG opt              :2026-05-22, 14d
    T14 Schema instructor    :2026-05-22, 7d
    T31 LLM test gen         :crit, 2026-05-22, 14d
    T8 Conv-outcome HubSpot  :crit, 2026-05-29, 14d
    T13 LLM-judge tuning     :2026-06-05, 14d

    section Wave 3 Advanced
    T3.3 TextGrad            :2026-06-12, 7d
    T4.1 Karpathy            :2026-06-12, 14d
    T4.2 Voyager             :2026-06-19, 21d
    T17 Constitutional       :2026-06-19, 14d
    T27 LLMLingua            :2026-06-26, 7d
    T30 Drift Evidently      :2026-06-26, 14d

    section Wave 4 Memory
    T5 Honcho/Mem0/Letta     :2026-07-10, 28d
    T34 Memory eviction      :2026-07-17, 14d
    T18 AutoGen/MetaGPT      :2026-07-17, 21d
    T35 MoA                  :2026-07-24, 14d
    T9 Tool selection MCP    :2026-07-31, 14d

    section Wave 5 Frontier
    T4.5 HALO eval           :2026-08-14, 28d
    T28 Test-time compute    :2026-08-14, 14d
    T29 PRM                  :2026-08-21, 21d
    Voice infra LiveKit      :crit, 2026-08-28, 21d

    section Wave 6 Voice
    T26 Voice patterns       :2026-09-18, 42d
    T22 Continual learning   :2026-10-09, 21d
    T16 Synthetic distill    :2026-10-30, 28d

    section Wave 7 Track-only
    Quarterly research scan  :2026-11-03, 90d

Graph View

Backlinks