Tool Use (Function Calling) & RAG (Slides)

# Day 2: Tool Use (Function Calling) + RAG

<div style="margin-top: 1.8em; text-align: center; opacity: 0.9;">
SCIPE Workshop on Large Language Models
</div>

<div style="margin-top: 2.0em; opacity: 0.75; font-size: 0.85em; text-align: center;">
Case Western Reserve University • January 17, 2026
</div>

---

.reveal .c-grid { display: grid; gap: 1.2rem; }
  .reveal .c-grid-2 { grid-template-columns: 1fr 1fr; }
  .reveal .c-grid-3 { grid-template-columns: 1fr 1fr 1fr; }
  .reveal .c-card {
    background: var(--card-bg);
    border: 1px solid var(--card-border);
    border-radius: 18px;
    padding: 1.15rem 1.25rem;
    box-shadow: 0 16px 34px rgba(15,23,42,0.07);
  }
  .reveal .c-kicker {
    font-size: 0.55em;
    letter-spacing: 0.14em;
    text-transform: uppercase;
    opacity: 0.7;
    margin-bottom: 0.4rem;
  }
  .reveal .c-muted { color: var(--muted); }
  .reveal .c-titleRow { display: flex; align-items: center; gap: 0.6rem; margin-bottom: 0.65rem; }
  .reveal .c-titleRow strong { font-weight: 700; }

.reveal .c-pill {
    display: inline-flex;
    align-items: center;
    gap: 0.4rem;
    padding: 0.25rem 0.55rem;
    border-radius: 999px;
    font-size: 0.55em;
    font-weight: 700;
    border: 1px solid rgba(15,23,42,0.14);
    background: rgba(255,255,255,0.75);
    letter-spacing: 0.04em;
    text-transform: uppercase;
  }
  .reveal .c-pill.tool { border-color: rgba(2,132,199,0.24); background: var(--tool-soft); color: rgba(2,132,199,1); }
  .reveal .c-pill.rag  { border-color: rgba(249,115,22,0.26); background: var(--rag-soft); color: rgba(194,65,12,1); }
  .reveal .c-pill.mix  { border-color: rgba(16,185,129,0.26); background: var(--mix-soft); color: rgba(5,150,105,1); }

.reveal .c-list { margin: 0.65rem 0 0 0; padding-left: 1.15rem; font-size: 0.82em; }
  .reveal .c-list li { margin: 0.28rem 0; }

.reveal .c-callout {
    border-left: 6px solid rgba(2,132,199,0.55);
    background: rgba(2,132,199,0.06);
    padding: 0.9rem 1.0rem;
    border-radius: 16px;
    font-size: 0.82em;
  }
  .reveal .c-callout.warn { border-left-color: rgba(249,115,22,0.70); background: rgba(249,115,22,0.08); }
  .reveal .c-callout.ok   { border-left-color: rgba(16,185,129,0.70); background: rgba(16,185,129,0.08); }

.reveal .c-flow { display: flex; align-items: stretch; justify-content: center; gap: 0.65rem; flex-wrap: wrap; margin-top: 1.0rem; }
  .reveal .c-node {
    min-width: 240px;
    padding: 0.85rem 0.95rem;
    border-radius: 16px;
    border: 1px solid rgba(15,23,42,0.12);
    background: rgba(255,255,255,0.72);
    text-align: center;
    font-size: 0.78em;
  }
  .reveal .c-node .small { display: block; margin-top: 0.35rem; font-size: 0.78em; opacity: 0.75; }
  .reveal .c-arrow { display: flex; align-items: center; font-size: 1.2em; opacity: 0.4; padding: 0 0.15rem; }

.reveal .c-pre {
    margin: 0.5rem 0 0 0;
    padding: 0.85rem 0.9rem;
    border-radius: 16px;
    background: rgba(15,23,42,0.05);
    border: 1px solid rgba(15,23,42,0.10);
    font-size: 0.68em;
    overflow: hidden;
  }
  .reveal .c-pre code { white-space: pre; }

.reveal .c-svg {
    width: 100%;
    height: auto;
    display: block;
    margin: 0 auto;
  }

.reveal .c-foot { font-size: 0.55em; opacity: 0.75; margin-top: 0.85rem; }

@media (max-width: 1100px) {
    .reveal .c-grid-3 { grid-template-columns: 1fr; }
    .reveal .c-grid-2 { grid-template-columns: 1fr; }
  }
</style>

## Today: from “chat” to “systems”

<div class="c-grid c-grid-3" style="margin-top: 1.1rem;">
  <div class="c-card">
    <div class="c-kicker">Part 1</div>
    <div class="c-titleRow">
      <span class="c-pill tool">Tools</span>
      <strong>Function calling</strong>
    </div>
    <div class="c-muted" style="font-size: 0.78em;">Turn natural language into structured, executable actions.</div>
    <ul class="c-list">
      <li>Tool schemas and validation</li>
      <li>Core tool-call loop</li>
      <li>Reliability patterns</li>
      <li>Security / trust boundaries</li>
    </ul>
  </div>

<div class="c-card">
    <div class="c-kicker">Part 2</div>
    <div class="c-titleRow">
      <span class="c-pill rag">RAG</span>
      <strong>Retrieval-augmented generation</strong>
    </div>
    <div class="c-muted" style="font-size: 0.78em;">Ground the model in your data, with citations.</div>
    <ul class="c-list">
      <li>Index-time vs query-time</li>
      <li>Vector search intuition</li>
      <li>Chunking, reranking, packing</li>
      <li>Failure modes & evaluation</li>
    </ul>
  </div>

<div class="c-card">
    <div class="c-kicker">Part 3</div>
    <div class="c-titleRow">
      <span class="c-pill mix">Combine</span>
      <strong>Agentic / tool-augmented RAG</strong>
    </div>
    <div class="c-muted" style="font-size: 0.78em;">Retrieval is “just another tool” — but needs guardrails.</div>
    <ul class="c-list">
      <li>Query rewriting tool</li>
      <li>Retrieval + rerank tool</li>
      <li>Grounded answer formatting</li>
      <li>Observability and caching</li>
    </ul>
  </div>
</div>

<div class="c-callout ok" style="margin-top: 1.2rem;">
  <strong>Goal:</strong> reliable answers grounded in data — plus actions that are auditable.
</div>

---

## LLM-only vs LLM-as-a-component

<div class="c-grid c-grid-2" style="margin-top: 1.1rem; align-items: start;">
  <div class="c-card">
    <div class="c-titleRow">
      <span class="c-pill" style="border-color: rgba(15,23,42,0.16); background: rgba(15,23,42,0.05); color: rgba(15,23,42,0.75);">Model only</span>
      <strong>Text in → text out</strong>
    </div>
    <ul class="c-list">
      <li>Can be wrong confidently (hallucinations)</li>
      <li>Can’t see private or fresh data</li>
      <li>Can’t take actions or verify outcomes</li>
      <li>Hard to audit “why this answer?”</li>
    </ul>
    <div class="c-callout warn" style="margin-top: 0.95rem;">
      <strong>Good for:</strong> brainstorming, drafting, summarizing what’s already provided.
    </div>
  </div>

<div class="c-card">
    <div class="c-titleRow">
      <span class="c-pill mix">System</span>
      <strong>LLM + tools + retrieval</strong>
    </div>
    <div class="c-flow" style="margin-top: 0.4rem;">
      <div class="c-node">
        User request
        <span class="small">natural language</span>
      </div>
      <div class="c-arrow">→</div>
      <div class="c-node">
        LLM
        <span class="small">decide + call</span>
      </div>
      <div class="c-arrow">→</div>
      <div class="c-node">
        Tools / RAG
        <span class="small">data + actions</span>
      </div>
      <div class="c-arrow">→</div>
      <div class="c-node">
        Grounded answer
        <span class="small">with evidence</span>
      </div>
    </div>
    <div class="c-callout ok" style="margin-top: 0.95rem;">
      <strong>Good for:</strong> production apps (search, analytics, workflows, assistants).
    </div>
  </div>
</div>

---

# Tool Use (Function Calling)

<div class="c-callout" style="max-width: 58rem; margin: 1.2rem auto;">
   the model emits <strong>structured arguments</strong> for a tool you own, your runtime validates them, executes, and returns <strong>structured results</strong>.
</div>

---

## Why tools?

<div class="c-grid c-grid-2" style="margin-top: 1.1rem; align-items: start;">
  <div class="c-card">
    <div class="c-titleRow">
      <span class="c-pill tool">Reason</span>
      <strong>LLMs don’t have guaranteed access to truth</strong>
    </div>
    <ul class="c-list">
      <li><strong>Freshness:</strong> news, prices, inventory, policies change</li>
      <li><strong>Private data:</strong> your docs, tickets, logs, databases</li>
      <li><strong>Determinism:</strong> math, code execution, ETL, search</li>
      <li><strong>Actions:</strong> create a ticket, send email, run workflow</li>
    </ul>
  </div>

<div class="c-card">
    <div class="c-titleRow">
      <span class="c-pill tool">Pattern</span>
      <strong>Turn “intent” into “API calls”</strong>
    </div>
    <div class="c-flow" style="margin-top: 0.2rem;">
      <div class="c-node">
        Intent
        <span class="small">“Book a room”</span>
      </div>
      <div class="c-arrow">→</div>
      <div class="c-node" style="border-color: rgba(2,132,199,0.22); background: rgba(2,132,199,0.06);">
        Tool call
        <span class="small"><code>create_reservation(...)</code></span>
      </div>
      <div class="c-arrow">→</div>
      <div class="c-node">
        Result
        <span class="small">confirmation + details</span>
      </div>
    </div>
    <div class="c-foot">
      Reliability comes from <strong>schemas</strong>, <strong>validation</strong>, and <strong>guardrails</strong>, not from “asking nicely”.
    </div>
  </div>
</div>

---

## The contract: schemas + validation

<div class="c-grid c-grid-2" style="margin-top: 1.0rem; align-items: start;">
  <div class="c-card">
    <div class="c-titleRow">
      <span class="c-pill tool">Tool</span>
      <strong>Definition (what the model can call)</strong>
    </div>
    <pre class="c-pre" style="font-size: 0.58em;"><code class="language-json">{
  "name": "search_docs",
  "description": "Search internal docs and return relevant chunks.",
  "parameters": {
    "type": "object",
    "properties": {
      "query": { "type": "string" },
      "top_k": { "type": "integer", "minimum": 1, "maximum": 10 }
    },
    "required": ["query"]
  }
}</code></pre>
    <div class="c-foot">
      Runtime responsibilities: <strong>validate types</strong>, enforce limits, handle auth/timeouts, log calls.
    </div>
  </div>

<div class="c-card">
    <div class="c-titleRow">
      <span class="c-pill tool">Call</span>
      <strong>Model emits structured arguments</strong>
    </div>
    <pre class="c-pre" style="font-size: 0.58em;"><code class="language-json">{
  "tool": "search_docs",
  "arguments": {
    "query": "What is retrieval-augmented generation?",
    "top_k": 5
  }
}</code></pre>
    <div class="c-callout ok" style="margin-top: 0.9rem;">
      <strong>Key idea:</strong> the model “writes JSON” — your system decides what gets executed.
    </div>
  </div>
</div>

---

## Core loop (tool calling)

<pre class="c-pre" style="margin-top: 1.0rem;"><code class="language-mermaid">%%{init: { 'theme': 'default', 'themeVariables': { 'fontSize': '26px' }, 'sequence': { 'messageFontSize': 24, 'noteFontSize': 22 }, 'config': { 'scale': 2 } }}%%
sequenceDiagram
  participant U as User
  participant M as LLM
  participant R as Runtime
  participant T as Tool/API

U->>M: Question + constraints
  M->>R: tool_call(name, args)
  R->>T: Execute tool(args)
  T-->>R: tool_result(data)
  R-->>M: tool_result(data)
  M-->>U: Final answer (grounded)</code></pre>

<div class="c-foot">
  The runtime is where <strong>permissions</strong>, <strong>validation</strong>, and <strong>auditing</strong> live.
</div>

---

## Agentic loop (plan → act → observe)

<div class="c-grid c-grid-2" style="margin-top: 1.1rem; align-items: center;">
  <div class="c-card">
    <div class="c-titleRow">
      <span class="c-pill tool">Loop</span>
      <strong>Multiple tool calls for harder tasks</strong>
    </div>
    <ul class="c-list">
      <li>decompose into steps</li>
      <li>call tools for each step</li>
      <li>update state (memory, scratchpad, citations)</li>
      <li>stop when success criteria met</li>
    </ul>
    <div class="c-callout warn" style="margin-top: 0.95rem;">
      <strong>Risk:</strong> more steps → more opportunities for mistakes and injection.
    </div>
  </div>

---

## Reliability patterns (what makes tools work in production)

<div class="c-grid c-grid-3" style="margin-top: 1.1rem;">
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill tool">Validate</span><strong>Strict schemas</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Reject / repair malformed args. Enforce bounds.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill tool">Retry</span><strong>Deterministic fallbacks</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Timeouts, retries, alternate tools, safe defaults.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill tool">Idempotent</span><strong>Safe side-effects</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Use request IDs; don’t double-charge / double-create.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill tool">Cache</span><strong>Memoize results</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Speed + cost control; cache by (tool, args, user).</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill tool">Trace</span><strong>Logs + spans</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Know which tool calls created which answers.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill tool">Guard</span><strong>Least privilege</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Allow-list tools; per-user auth; redact secrets.</div>
  </div>
</div>

---

## Security: trust boundaries (prompt injection is a systems problem)

<div class="c-grid c-grid-2" style="margin-top: 1.1rem; align-items: start;">
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill tool">Threat</span><strong>Untrusted inputs everywhere</strong></div>
    <ul class="c-list">
      <li>user prompt can try to override system goals</li>
      <li>tool outputs can contain adversarial text (“ignore prior instructions…”)</li>
      <li>retrieved documents can smuggle instructions (RAG injection)</li>
    </ul>
    <div class="c-callout warn" style="margin-top: 0.95rem;">
      <strong>Rule:</strong> treat tool output and retrieved text as <strong>data</strong>, not instructions.
    </div>
  </div>

---

# Retrieval-Augmented Generation (RAG)

<div class="c-callout" style="border-left-color: rgba(249,115,22,0.75); background: rgba(249,115,22,0.07); max-width: 58rem; margin: 1.2rem auto;">
   retrieve the most relevant chunks from your corpus and feed them to the model as context.
</div>

---

## RAG: Big Picture

<div class="c-grid c-grid-2" style="margin-top: 1.1rem; align-items: start;">
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">Index time</span><strong>Build the knowledge base</strong></div>
    <ul class="c-list">
      <li>collect documents (PDF, HTML, wiki, tickets)</li>
      <li>chunk (size + overlap) and add metadata</li>
      <li>embed chunks into vectors</li>
      <li>store in a vector index</li>
    </ul>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">Query time</span><strong>Answer with evidence</strong></div>
    <ul class="c-list">
      <li>embed the user query</li>
      <li>retrieve top-k chunks (vector / hybrid)</li>
      <li>rerank + pack into context window</li>
      <li>generate an answer with citations</li>
    </ul>
  </div>
</div>

<div class="c-foot">
  RAG is a <strong>data pipeline</strong> and a <strong>prompting pattern</strong> — not a single library call.
</div>

---

## RAG pipeline (end-to-end)

<pre class="c-pre" style="margin-top: 1.0rem;"><code class="language-mermaid">%%{init: { 'theme': 'default', 'themeVariables': { 'fontSize': '24px' }, 'flowchart': { 'nodeSpacing': 50, 'rankSpacing': 55 }, 'config': { 'scale': 2 } }}%%
flowchart LR
  subgraph I[Index time]
    D[Docs] --> C[Chunk + metadata]
    C --> E[Embed]
    E --> V[(Vector index)]
  end

subgraph Q[Query time]
    U[User question] --> QE[Query embed]
    QE --> R[Retrieve top-k]
    V --> R
    R --> RR[Rerank / filter]
    RR --> P[Pack context]
    P --> L[LLM]
    L --> A[Answer + citations]
  end</code></pre>

---

## Vector search intuition

<div class="c-grid c-grid-2" style="margin-top: 1.1rem; align-items: center;">
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">Idea</span><strong>Nearest neighbors in embedding space</strong></div>
    <ul class="c-list">
      <li>each chunk becomes a vector</li>
      <li>query becomes a vector</li>
      <li>retrieve chunks closest to the query vector</li>
    </ul>
    <div class="c-callout ok" style="margin-top: 0.95rem;">
       mix semantic search with keyword filters (hybrid) and reranking.
    </div>
  </div>

---

## Context packing (why “just add more docs” fails)

<div class="c-grid c-grid-2" style="margin-top: 1.1rem; align-items: start;">
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">Constraint</span><strong>Context window is finite</strong></div>
    <ul class="c-list">
      <li>too much context → relevant pieces get diluted</li>
      <li>too little context → missing evidence</li>
      <li>ordering matters (high-signal first)</li>
    </ul>
    <div class="c-callout warn" style="margin-top: 0.95rem;">
      <strong>Practical rule:</strong> retrieve, filter, rerank, then pack — don’t “dump”.
    </div>
  </div>

<div class="c-card">
    <div style="font-size: 0.75em; opacity: 0.75; margin-bottom: 0.5rem;">Context window (tokens)</div>
    <div style="display: flex; height: 46px; border-radius: 14px; overflow: hidden; border: 1px solid rgba(15,23,42,0.12); background: rgba(255,255,255,0.65);">
      <div style="flex: 3; background: rgba(249,115,22,0.22); border-right: 1px solid rgba(15,23,42,0.10); display: flex; align-items: center; justify-content: center; font-size: 0.68em; font-weight: 700; color: rgba(194,65,12,0.95);">Top chunks</div>
      <div style="flex: 2; background: rgba(2,132,199,0.12); border-right: 1px solid rgba(15,23,42,0.10); display: flex; align-items: center; justify-content: center; font-size: 0.68em; font-weight: 700; color: rgba(2,132,199,0.95);">Definitions</div>
      <div style="flex: 2; background: rgba(16,185,129,0.12); border-right: 1px solid rgba(15,23,42,0.10); display: flex; align-items: center; justify-content: center; font-size: 0.68em; font-weight: 700; color: rgba(5,150,105,0.95);">Constraints</div>
      <div style="flex: 1; background: rgba(15,23,42,0.06); display: flex; align-items: center; justify-content: center; font-size: 0.68em; font-weight: 700; color: rgba(15,23,42,0.70);">…</div>
    </div>
    <div class="c-foot">
      Good packers use <strong>dedup</strong>, <strong>MMR</strong>, and <strong>metadata filters</strong> (recency, source).
    </div>
  </div>
</div>

---

## Failure modes (and the fix is rarely “bigger model”)

<div class="c-grid c-grid-2" style="margin-top: 1.1rem;">
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">Miss</span><strong>No relevant chunks retrieved</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Fix: better chunking, query rewrite, hybrid retrieval, metadata filters.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">Buried</span><strong>Relevant chunk is low-ranked</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Fix: reranking (cross-encoder), MMR, per-section retrieval.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">Drift</span><strong>Answer ignores evidence</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Fix: stronger grounding prompt, citations required, constrained format.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">Injection</span><strong>Retrieved text attacks the model</strong></div>
    <div class="c-muted" style="font-size: 0.78em;">Fix: isolate retrieved content, strip instructions, run safety filters.</div>
  </div>
</div>

<!-- ---

## Evaluation checklist

<div class="c-card" style="margin-top: 1.1rem;">
  <div class="c-titleRow"><span class="c-pill mix">Measure</span><strong>Retrieval + generation + cost</strong></div>
  <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 1.0rem; margin-top: 0.8rem; font-size: 0.78em;">
    <div style="padding: 0.85rem 0.95rem; border-radius: 16px; border: 1px solid rgba(15,23,42,0.10); background: rgba(255,255,255,0.65);">
      <div style="font-weight: 800; margin-bottom: 0.4rem;">Retrieval quality</div>
      <div class="c-muted">Recall@k, MRR, nDCG, source coverage, filter accuracy</div>
    </div>
    <div style="padding: 0.85rem 0.95rem; border-radius: 16px; border: 1px solid rgba(15,23,42,0.10); background: rgba(255,255,255,0.65);">
      <div style="font-weight: 800; margin-bottom: 0.4rem;">Answer quality</div>
      <div class="c-muted">Correctness, faithfulness, citation correctness, refusal when unknown</div>
    </div>
    <div style="padding: 0.85rem 0.95rem; border-radius: 16px; border: 1px solid rgba(15,23,42,0.10); background: rgba(255,255,255,0.65);">
      <div style="font-weight: 800; margin-bottom: 0.4rem;">Latency & cost</div>
      <div class="c-muted">p50/p95 latency, tokens, tool costs, cache hit rate</div>
    </div>
    <div style="padding: 0.85rem 0.95rem; border-radius: 16px; border: 1px solid rgba(15,23,42,0.10); background: rgba(255,255,255,0.65);">
      <div style="font-weight: 800; margin-bottom: 0.4rem;">Safety</div>
      <div class="c-muted">Injection robustness, data leakage, policy adherence, least privilege</div>
    </div>
  </div>
</div> -->

---

# Tool‑Augmented RAG

<div class="c-callout ok" style="max-width: 58rem; margin: 1.2rem auto;">
  <strong>Idea:</strong> treat retrieval as one tool among many — and let the agent decide when to retrieve, rewrite, verify, and cite.
</div>

---

## Agentic RAG (one practical architecture)

<pre class="c-pre" style="margin-top: 1.0rem;"><code class="language-mermaid">%%{init: { 'theme': 'default', 'themeVariables': { 'fontSize': '22px' }, 'flowchart': { 'nodeSpacing': 40, 'rankSpacing': 50 }, 'config': { 'scale': 2 } }}%%
flowchart TD
  U[User question] --> M[LLM]
  M -->|rewrite| Q[Query rewrite tool]
  Q --> R[Retriever tool]
  R --> RR[Reranker / filters]
  RR --> P[Context packer]
  P --> M
  M -->|optional| V[Verifier tool]
  V --> M
  M --> A[Answer + citations]</code></pre>

<div class="c-foot">
  The win is <strong>selective context</strong> + <strong>audited steps</strong> — the risk is uncontrolled loops.
</div>

---

## Key takeaways

<div class="c-grid c-grid-2" style="margin-top: 1.1rem;">
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill tool">Tools</span><strong>Reliability comes from engineering</strong></div>
    <div class="c-muted" style="font-size: 0.80em;">Schemas, validation, auth, retries, tracing, and least privilege.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill rag">RAG</span><strong>Grounding is a pipeline</strong></div>
    <div class="c-muted" style="font-size: 0.80em;">Chunking + retrieval + reranking + packing + prompting + evaluation.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill mix">Combine</span><strong>Retrieval is a tool call</strong></div>
    <div class="c-muted" style="font-size: 0.80em;">Agentic RAG is powerful when guarded by runtime policy.</div>
  </div>
  <div class="c-card">
    <div class="c-titleRow"><span class="c-pill mix">Next</span><strong>Hands-on</strong></div>
    <div class="c-muted" style="font-size: 0.80em;">Build a small app</div>
  </div>
</div>

---

# Thank You!

<h3 style="margin-bottom: 2rem; font-size: 2em;">Questions?</h3>

<div style="font-size: 1.5em; font-weight: bold; text-align: center; margin-top: 2rem; opacity: 0.95;">
Now let's start coding!
</div>

</div>

Navigation