Stacks / 7 workflow stages · v3.1.0 · reviewed 2026-09-05
Build an Internal Knowledge Assistant
Ingest documents, build retrieval, produce grounded answers, and test whether the system actually retrieves the right evidence.
Best for
Teams building an internal assistant over company documents, policies, research, product knowledge, or other private information.
Outcome
A grounded knowledge workflow with ingestion, retrieval/orchestration, structured answers, evaluation, and observability.
Not for
A public-web research agent whose primary job is browsing and discovering new external information.
Implementation noteThe current Index does not include a strong general-purpose vector database resource. Keep the organization’s approved database/vector layer outside this Stack rather than implying LlamaIndex replaces persistence.
Operator guidance
Best whenAnswers must stay grounded in changing private content with traceable citations.
Avoid whenThe corpus is tiny, static, or better served by deterministic search.
CompatibilityUse one metadata and document-ID contract across ingestion, retrieval, generation, evaluation, and observability.
Use this as a decision framework, not a mandatory shopping list. Swap or skip layers based on your existing stack, constraints, and risk profile.
1. Prepare content
Convert web and document-like content into cleaner text that can be indexed and reasoned over.
- Firecrawl
Recommendation details
Why: Consider Firecrawl as an alternative for 1. Prepare content; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Web extraction and crawling for agent workflows.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Firecrawl official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Convert web pages into LLM-friendly text.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Jina AI Reader official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
2. Build retrieval
Use a framework centered on data ingestion, indexing, retrieval, and knowledge-grounded applications.
- LangGraph
Recommendation details
Why: Consider LangGraph as an alternative for 2. Build retrieval; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Stateful agent/workflow framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: LangGraph official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Retrieval and agent framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: LlamaIndex official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
3. Structure outputs
Constrain important answers into predictable structured outputs when downstream systems need fields rather than prose.
- Pydantic AI
Recommendation details
Why: Consider Pydantic AI as an alternative for 3. Structure outputs; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Typed agent framework from Pydantic.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Pydantic AI official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Reliable structured extraction with LLMs.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Instructor official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
4. Add workflow control
Use stateful orchestration when retrieval needs routing, retries, multiple sources, or tool use.
- Dify
Recommendation details
Why: Consider Dify as an alternative for 4. Add workflow control; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Production-ready platform for agentic workflows, RAG, model management and observability.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Dify official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Stateful agent/workflow framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: LangGraph official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
5. Evaluate retrieval
Measure retrieval and answer quality rather than assuming a RAG system is grounded because it returns citations.
- DeepEval
Recommendation details
Why: Consider DeepEval as an alternative for 5. Evaluate retrieval; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Code-first evaluation framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: DeepEval official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- Braintrust
Recommendation details
Why: Consider Braintrust as an alternative for 5. Evaluate retrieval; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Evals, traces and production quality workflows.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Braintrust official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: RAG-focused evaluation framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Ragas official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
6. Observe production
Trace retrieval, model calls, and failures after deployment so weak answers can be diagnosed.
- Arize Phoenix
Recommendation details
Why: Consider Arize Phoenix as an alternative for 6. Observe production; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Open-source tracing and evaluation.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Arize Phoenix official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- LangSmith
Recommendation details
Why: Consider LangSmith as an alternative for 6. Observe production; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Tracing and observability platform for agent and LLM applications.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: LangSmith official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Open-source LLM observability and evals.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Langfuse official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
7. Review security
Check prompt injection, data leakage, access-control, and other risks that matter when an assistant touches internal information.
- Anthropic Security Guidance
Recommendation details
Why: Consider Anthropic Security Guidance as an alternative for 7. Review security; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Anthropic plugin with edit warnings and diff review for common vulnerability classes.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Anthropic Security Guidance official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: High-value LLM risk checklist.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: OWASP Top 10 for LLM Applications official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Implementation references
-
Vercel Chat SDK Knowledge Agent ↗
A compact, working knowledge-agent reference with file-system grounding and a user-facing chat surface.
vercel.com · template -
Eugene Yan: Patterns for Building LLM Systems ↗
Production patterns for retrieval, evaluation, defensive design, caching, and observability.
eugeneyan.com · reference architecture
Need this workflow built and running, not just linked? EE Solutions implements agentic stacks like this one. EE Solutions is a senior technology team for private capital firms and their portfolio companies.
Talk to EE Solutions ↗