Practical notes: Semantic Caching for AI Agents: How to Make Your LLM Apps

Use this as an operator-facing rebuild of the ideas in “Semantic Caching for AI Agents: How to Make Your LLM Apps Faster and Cheaper”: clear stages, ordered code slots, and recovery notes that survive a handoff. The Overview stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

How Semantic Caching Works

For the How Semantic Caching Works stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Building a Semantic Cache from Scratch

For the Building a Semantic Cache stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

from sentence_transformers import SentenceTransformer
import numpy as np

model = SentenceTransformer("all-mpnet-base-v2")# Embed your FAQ dataset
faq_questions = ["How do I get a refund?", "Where is my order?", ...]
faq_embeddings = model.encode(faq_questions)def cosine_distance(a, b):
    return 1 - np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b))def check_cache(query: str, threshold: float = 0.3):
    query_embedding = model.encode(query)
    distances = [cosine_distance(query_embedding, e) for e in faq_embeddings]
    best_idx = np.argmin(distances)
    best_distance = distances[best_idx]

    if best_distance < threshold:
        return faq_answers[best_idx]  # Cache hit
    return None  # Cache miss

Moving to Production with Redis

For the Moving to Production with stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call. For the Moving to Production with stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

from redisvl.extensions.cache.llm import SemanticCache
from redisvl.utils.vectorize import HFTextVectorizer

# Load a cache-optimized embedding model
vectorizer = HFTextVectorizer("redis/langcache-embed-v1")# Create the cache
cache = SemanticCache(
    name="customer_support_cache",
    vectorizer=vectorizer,
    redis_client=redis_client,
    distance_threshold=0.3
)# Set TTL (time to live) — keeps cache fresh
cache.set_ttl(86400)  # 24 hours

Measuring What You’ve Built

When working through the Measuring What You ve stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn.

1. Cache Hit Rate

When working through the 1 Cache Hit Rate stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn.

2. Precision

When working through the 2 Precision stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn. When working through the 2 Precision stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

3. Recall

The 3 Recall stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices.

4. Latency Improvement

The 4 Latency Improvement stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices.

With-Cache Latency = (avg_llm_latency × (1 - hit_rate)) + (avg_cache_latency × hit_rate)

The Confusion Matrix View

The The Confusion Matrix View stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices. The The Confusion Matrix View stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Four Techniques to Improve Cache Accuracy

For the Four Techniques to Improve stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

1. Threshold Sweep

For the 1 Threshold Sweep stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

2. Cross-Encoder Reranking

For the 2 Cross-Encoder Reranking stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call. For the 2 Cross-Encoder Reranking stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

3. LLM-as-a-Judge

When working through the 3 LLM-as-a-Judge stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn.

4. Fuzzy Matching as a Pre-Filter

When working through the 4 Fuzzy Matching as stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn.

The Real Payoff: Caching Inside AI Agents

When working through the The Real Payoff Caching stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn. When working through the The Real Payoff Caching stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Real-World Validation: Walmart’s waLLMartCache

The Real-World Validation Walmart s stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices.

Getting Started

The Getting Started stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices.

Conclusion

The Conclusion stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Budget tokens per turn and per session. Agentic tools expand context aggressively; hard caps keep demos from becoming surprise invoices. The Conclusion stage works best when treated as a measurable surface. Capture one golden transcript, one failure case, and the rollback note before expanding scope. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion.

Operational checklist

When working through the Operational checklist stage, write down the contract first: required inputs, success signal, and what happens on partial failure. That checklist keeps later code changes honest.

Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline.

Cache stable system instructions and tool schemas. Re-sending identical preamble is a common source of burn.

Keep graph state flat and typed. Nested blobs hide which node wrote which field and break resume after interrupts.

Add a smoke test that exercises the critical path in CI with fixtures, not live paid APIs, whenever budgets allow.

Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph.

Before promoting the stack, freeze versions, capture a golden transcript for the critical path, and confirm rollback steps. Shared environments need rate limits, tenancy checks, and a clear owner for secret rotation. Prefer boring reliability over clever one-off demos.

Batch note for 576e7f2969bf: keep provider keys out of the repo, set a per-session token ceiling, and store transcripts next to the eval fixtures so later model swaps stay comparable.