Embedding Strategies
Software development skill, available on Zeplik
Embedding Strategies is a ready-to-run software development skill on Zeplik. Not for building a full RAG pipeline (use rag-implementation). Ask in plain language and Zeplik applies the skill's method for you inside the conversation, on whichever AI model you prefer.
The Embedding Strategies skill loads automatically when your request matches it, or you can invoke it directly by typing /embedding-strategies in any chat. It works with attachments, connectors, and any model that supports the task, so you get the same expert method every time without setting anything up.
What the Embedding Strategies skill can do
- Recommend embedding models based on domain, cost, and latency needs
- Design chunking strategies matched to token limits and query granularity
- Compare embedding models on dimensions, context length, and use case fit
- Propose small-scale recall@k evaluations to validate model choice
Try these prompts on Zeplik
Pick a prompt to open it in the Zeplik app. If you are not signed in yet, your prompt is waiting for you the moment you do.
How the Embedding Strategies skill works
/embedding-strategies
Select and optimize embedding models for vector search applications. The user describes their corpus, domain, and constraints (or pastes sample documents and current chunking code); deliver a model recommendation with rationale, a chunking design, and example pipeline code as chat artifacts. For assembling the end-to-end retrieval-plus-generation system, use rag-implementation.
When to Use
- Choosing embedding models for RAG
- Optimizing chunking strategies
- Fine-tuning embeddings for domains
- Comparing embedding model performance
- Reducing embedding dimensions
- Handling multilingual content
Core Concepts
1. Embedding Model Comparison (2026)
| Model | Dimensions | Max Tokens | Best For |
|---|---|---|---|
| voyage-3-large | 1024 | 32000 | Claude apps (Anthropic recommended) |
| voyage-3 | 1024 | 32000 | Claude apps, cost-effective |
| voyage-code-3 | 1024 | 32000 | Code search |
| voyage-finance-2 | 1024 | 32000 | Financial documents |
| voyage-law-2 | 1024 | 32000 | Legal documents |
| text-embedding-3-large | 3072 | 8191 | OpenAI apps, high accuracy |
| text-embedding-3-small | 1536 | 8191 | OpenAI apps, cost-effective |
| bge-large-en-v1.5 | 1024 | 512 | Open source, local deployment |
| all-MiniLM-L6-v2 | 384 | 256 | Fast, lightweight |
| multilingual-e5-large | 1024 | 512 | Multi-language |
2. Embedding Pipeline
Document -> Chunking -> Preprocessing -> Embedding Model -> Vector
| | |
[Overlap, Size] [Clean, Normalize] [API/Local]
Recommendation Workflow
- Ask for the corpus profile: domain, language(s), document lengths, query style, latency and cost constraints, hosting requirements (API allowed vs local-only).
- Shortlist 2-3 models from the table with explicit trade-offs.
- Design chunking: size, overlap, and boundary strategy (headings, sentences, code blocks) matched to the model's token limit and the queries' granularity.
- Propose a small offline evaluation: a handful of representative queries with expected passages, measured via recall@k on each candidate model.
- Deliver the recommendation memo and pipeline snippet as chat artifacts.
Best Practices
Do's
- Match model to use case: Code vs prose vs multilingual
- Chunk thoughtfully: Preserve semantic boundaries
- Normalize embeddings: For cosine similarity search
- Batch requests: More efficient than one-by-one
- Cache embeddings: Avoid recomputing for static content
- Use Voyage AI for Claude apps: Recommended by Anthropic
Don'ts
- Don't ignore token limits: Truncation loses information
- Don't mix embedding models: Incompatible vector spaces
- Don't skip preprocessing: Garbage in, garbage out
- Don't over-chunk: Lose important context
- Don't forget metadata: Essential for filtering and debugging
Templates and Detailed Worked Examples
The full template library (chunking implementations, batching and caching code, domain fine-tuning, dimension reduction) lives in references/details.md. Read that file when you need the concrete templates.
Usage
/embedding-strategies $ARGUMENTS
How to use the Embedding Strategies skill
Sign in to Zeplik
Create a free Zeplik account or sign in. New accounts start with free credits, so you can try the Embedding Strategies skill right away.
Describe your software development task
Ask in plain language, or type /embedding-strategies to invoke the skill directly. Zeplik recognizes the Embedding Strategies skill and applies its method.
Review and refine the result
Zeplik returns a clear, structured answer. Ask follow-ups in the same chat to refine it or take the next step.
Source and credit
- Author
- wshobson
- License
- MIT
Adapted from the open-source wshobson/agents project and tuned to run natively on Zeplik. View source on GitHub.
Frequently asked questions
- What is the Embedding Strategies skill?
- Embedding Strategies is a ready-to-run software development skill on Zeplik. Not for building a full RAG pipeline (use rag-implementation). Ask in plain language and Zeplik applies the skill's method for you inside the conversation, on whichever AI model you prefer.
- How do I use Embedding Strategies on Zeplik?
- Sign in to Zeplik and ask in plain language, or type /embedding-strategies in any chat to invoke it directly. The skill applies its method and returns a result you can refine in the same conversation.
- Which AI model does the Embedding Strategies skill use?
- Any model you choose. Zeplik works across every model in one chat, so the Embedding Strategies skill runs on your preferred model for the task.
- Where does the Embedding Strategies skill come from?
- The Embedding Strategies skill is adapted from the open-source wshobson/agents project (MIT) and tuned to run natively on Zeplik. The original source is linked on this page.
- How much does the Embedding Strategies skill cost?
- Using the skill is free to start. You only spend Zeplik credits when the assistant runs, and new accounts begin with free credits.
Related software development skills
- .NET BackendBuild ASP.NET Core 8+ backends with EF Core: auth, background jobs, production API patterns
- Advanced Git WorkflowsUse for advanced Git surgery: interactive rebase, cherry-pick, bisect, reflog recovery, and history cleanup before merging. Not for parallel worktree workflows (use using-git-worktrees).
- Adversarial Code ReviewHunt for bugs in code the user shares by assuming defects exist and attacking the code through several distinct lenses, then report severity-ranked findings with evidence. Use for "review this", "what could go wrong", "bug hunt", or pre-merge scrutiny of a change. Read-only, it reports problems and does not rewrite the code. Not for style cleanup (use simplify-code) or for writing new code.
- AI Agent FrameworksUse when building multi-agent systems or agent orchestration -- LangChain/LangGraph, agent team design, task coordination, pipelines. Not for authoring a Zeplik skill (use skill-creator).
- Algolia SearchAdd Algolia search: indexing strategies, React InstantSearch, relevance tuning, search-as-you-type
- Android CI/CDAutomate Android CI/CD to Google Play: keystore, GitHub Secrets, multi-stage release workflow for RN, Flutter, native
More on Zeplik
Try Embedding Strategies on Zeplik
Every model, one chat. Bring the Embedding Strategies skill into your next conversation and let the assistant do the work.