AI Ecosystem KB

Seed reference map for the modern AI ecosystem, organized by stack layer and connected to concept/entity pages.

Full series Previous: AI Ecosystem Coverage Plan Next: Agent Development Tools Inventory

Introduction

AI Ecosystem KB is a reference page in the LLM Wiki, so it is designed to preserve source-backed facts, scope decisions, and maintenance cues that other articles can depend on. Seed reference map for the modern AI ecosystem, organized by stack layer and connected to concept/entity pages. The introduction explains what the reference covers, why it belongs in a durable wiki page, and how readers should use it when updating or validating the broader AI ecosystem map.

Reference pages are most valuable when they make their boundaries visible. The terms https, docs, documentation, microsoft, models, embeddings describe the evidence and classification surface for this page, while the body should show where the information came from, what has already been organized, and what still needs refresh. Readers should treat this article as a maintenance anchor: a place to verify coverage, connect related pages, and avoid duplicating source notes across the vault.

The AI ecosystem KB is the compiled map of modern AI tooling in this vault. It starts from the ai-ecosystem skill as a seed taxonomy, then organizes the landscape into durable wiki pages rather than a single long reference list.

Taxonomy Infographic

The taxonomy infographic groups representative tools by ecosystem layer and is optimized for scanning the landscape. It uses text badges instead of vendor logos so the asset remains self-contained inside the vault. It is curated rather than exhaustive; fast-moving categories should be refreshed against official sources before publication-grade use.

The KB should be maintained as a graph:

• Core layers live as concept pages: Foundation Models, Agent Orchestration, Retrieval-Augmented Generation, Embedding Layer, Model Context Protocol, AI Security Guardrails Governance, AI Observability Evaluation, AI Memory Management, Agent Development Frameworks, Workflow Automation Orchestration, and Vector Databases.
• Concrete vendors, frameworks, and products live as entity pages only when they become important enough to compare, choose, deploy, or monitor.
• Cross-layer decisions belong in synthesis pages such as AI Ecosystem Stack Patterns.

Update Policy

The ecosystem changes quickly. Product names, model families, context windows, pricing, lifecycle status, and availability should be verified against official vendor documentation before being written as current fact.

The seed skill is useful for layer structure and initial tool inventory. It should not be treated as authoritative for fast-moving model versions or vendor claims without a freshness check. ^[inferred]

Layer Map

Poster Category	Concept Page	Scope
LLM	Foundation Models	Model providers, open-weight models, local runners, and inference serving engines
Agentic AI	Agent Orchestration	Stateful planning, multi-agent coordination, handoffs, loops, and workflow graphs
RAG	Retrieval-Augmented Generation	Frameworks and pipelines for grounding model output in external knowledge
Embedding	Embedding Layer	Embedding models and services that turn content into semantic vectors
MCP	Model Context Protocol	Protocols and servers that expose tools, files, databases, and SaaS systems to AI apps
AI Security	AI Security Guardrails Governance	Guardrails, PII handling, prompt-injection defense, safety filters, and governance controls
Observability	AI Observability Evaluation	Tracing, evals, monitoring, cost, latency, red-team tests, and quality gates
Memory	AI Memory Management	Persistent user, task, session, episodic, semantic, relational, and short-term memory
AI Agent	Agent Development Frameworks	SDKs and managed platforms for defining, running, and deploying agent applications
Automation	Workflow Automation Orchestration	Low-code and code-first automation, integration, scheduling, and durable workflow execution
Vector Database	Vector Databases	Vector stores, hybrid search engines, and database-native vector search capabilities

Sources

This source catalog favors official documentation and project homepages. It supports the taxonomy and gives future refreshes a starting point; it is not a claim that every listed product is feature-equivalent or in the same maturity tier.

Seed Source

• Local ai-ecosystem skill taxonomy seed — seed taxonomy and initial tool inventory used to bootstrap this reference map inside the LLM Wiki.

Foundation Models and Serving

• OpenAI model documentation
• Anthropic Claude model overview
• Google Gemini API model documentation
• Meta Llama
• Mistral AI models
• Cohere models
• Hugging Face Models
• Ollama model library
• vLLM documentation
• DeepSeek API documentation
• xAI Grok API documentation
• Amazon Nova in Amazon Bedrock

Agentic AI and Agent Frameworks

• LangGraph documentation
• CrewAI documentation
• Microsoft AutoGen documentation
• Microsoft Agent Framework documentation
• LlamaIndex Workflows
• AWS Strands Agents
• CAMEL documentation
• Agno documentation
• OpenAI Agents SDK
• LangChain agents
• PydanticAI documentation
• Microsoft Semantic Kernel
• Google Agent Development Kit
• AWS Bedrock Agents
• Azure AI Foundry Agent Service

RAG, Parsing, and Knowledge Workflows

• LangChain documentation
• LlamaIndex documentation
• Haystack documentation
• DSPy documentation
• RAGFlow documentation
• Microsoft GraphRAG
• Unstructured documentation
• Embedchain documentation

Embeddings

• OpenAI embeddings documentation
• Cohere Embed
• Voyage AI embeddings
• Sentence Transformers
• BAAI BGE models on Hugging Face
• Google Vertex AI embeddings
• Azure OpenAI embeddings
• Jina embeddings
• Nomic Embed
• Mistral embeddings

Model Context Protocol

• Model Context Protocol documentation
• MCP SDKs
• FastMCP
• MCP servers repository
• GitHub MCP Server

Security, Guardrails, and Governance

• NVIDIA NeMo Guardrails
• Guardrails AI documentation
• Microsoft Presidio
• Lakera Guard
• Prompt Security
• Protect AI
• Azure AI Content Safety
• Amazon Bedrock Guardrails
• Meta Prompt Guard
• LlamaFirewall

Observability and Evaluation

• LangSmith documentation
• Langfuse documentation
• Arize Phoenix documentation
• Weights & Biases Weave
• TruLens documentation
• Ragas documentation
• Promptfoo documentation
• Helicone documentation
• Braintrust documentation
• Galileo documentation
• OpenTelemetry documentation

Memory and Datastores

• Mem0 documentation
• Zep documentation
• Letta documentation
• LangGraph memory
• Redis documentation
• PostgreSQL documentation
• Neo4j documentation
• Chroma documentation

Automation and Workflow Orchestration

• n8n documentation
• Zapier developer documentation
• Make documentation
• Microsoft Power Automate documentation
• Temporal documentation
• Apache Airflow documentation
• Prefect documentation
• Kestra documentation
• Pipedream documentation

Vector Databases and Search

• Pinecone documentation
• Weaviate documentation
• Qdrant documentation
• Milvus documentation
• Chroma documentation
• pgvector
• Elasticsearch vector search
• Redis vector search
• MongoDB Atlas Vector Search
• LanceDB documentation
• OpenSearch vector search
• Vespa documentation
• DataStax Astra DB vector search

Practical Implementation Context

For the AI Ecosystem KB reference page, practical implementation means maintaining a taxonomy source map that keeps source-backed knowledge reusable. The page should make scope, evidence, update signals, and downstream page ownership explicit so an editor can answer where a new AI ecosystem fact belongs without rediscovering the source trail.

• Maintain the page around layer boundary, official source group, and refresh owner.
• Trigger updates when concept page link, inventory path, or volatile claim changes.
• Use the workflow: check source group, update layer map, then link affected pages.
• Consider the reference healthy when new source material lands in the right concept, inventory, or entity page.

Implementation note: use this taxonomy source map to route updates from concept page link, inventory path, and volatile claim into the right concept, inventory, entity, or synthesis page.

Reference Implementation Pattern

For the AI Ecosystem KB reference page, the implementation pattern is a taxonomy source map. The page should make scope, source groups, freshness policy, and open update points explicit so the wiki can answer where a new AI ecosystem fact belongs without re-reading the whole vault.

reference_page:
  title: "AI Ecosystem KB"
  artifact: "taxonomy source map"
  decision_supported: "where a new AI ecosystem fact belongs"
  source_inbox: "_raw/ai-ecosystem-kb/"
  maintenance_checks:
    - "layer boundary"
    - "official source group"
    - "refresh owner"
  update_signals:
    - "concept page link"
    - "inventory path"
    - "volatile claim"

A concrete workflow is to check source group, update layer map, and link affected pages. The update is successful when new source material lands in the right concept, inventory, or entity page, and when related concept, inventory, entity, or synthesis links still point to the right page.

Key Takeaways

• Treat the source page as distilled knowledge, then add enough implementation context for a standalone reader.
• Make trade-offs visible: reliability, observability, governance, cost, and maintenance burden all matter.
• Use structured headings, tables, examples, and explicit warnings to help readers scan and apply the material.

Operational Depth

Source Stewardship

AI Ecosystem KB reference page should operate as a taxonomy source map. It preserves evidence and scope boundaries so the vault can answer where a new AI ecosystem fact belongs without rediscovering the source trail.

Refresh Signals

Operational review should inspect layer boundary, official source group, and refresh owner; update pressure usually appears as concept page link, inventory path, or volatile claim.

Validation Run

The reference page is healthy when a maintainer can check source group, update layer map, and link affected pages. The proof point is that new source material lands in the right concept, inventory, or entity page.

Review Cadence

Review this page whenever source material changes, linked pages are promoted, or a reader would make a different decision because of new information. The review should check content accuracy, link integrity, and whether the operational proof still matches the current LLM Wiki graph.

Reader Outcome

A reader should know what facts are stable, what needs refresh, and where updates should flow next.

Frequently Asked Questions

What job does AI Ecosystem KB perform in the wiki?

It keeps source-backed facts, scope notes, and refresh cues in one place so other pages can depend on a stable reference layer.

How should this reference page be maintained?

Review official source links, check volatile claims, update related links, and use ingest or lint workflows when the page needs new knowledge or structural cleanup.

When should reference content move elsewhere?

Move repeated patterns into concepts, important tools into entities, comparison material into inventories, and decision guidance into synthesis pages.

Conclusion

AI Ecosystem KB closes as a maintenance anchor for the LLM Wiki. Its purpose is to keep source-backed facts, coverage boundaries, refresh notes, and related links in one place so readers and agents can update the knowledge graph without duplicating evidence across many pages.

The most useful next step is to treat https, docs, documentation, microsoft as refresh cues: check whether the underlying sources are still current, whether missing pages should be created, and whether any claims should be promoted into concepts, entities, inventories, or synthesis articles. A good reference page makes future updates cheaper.

Full series Previous: AI Ecosystem Coverage Plan Next: Agent Development Tools Inventory