Skip to content

GitLab

"docs/design-docs/request-plane.md" did not exist on "dd6c399565fe203898e14f1d92c87be35f07f24f"
Unverified Commit 218cb13c authored by akshatha-k's avatar akshatha-k Committed by GitHub
Browse files

docs: add Agentic Workloads feature guide (#6982) 


Signed-off-by: default avatarakshatha-k <akshutk@gmail.com>
Signed-off-by: default avatarakshatha-k <33278067+akshatha-k@users.noreply.github.com>
Co-authored-by: default avatarBryan Bednarski <bbednarski@nvidia.com>
parent f435da14
Show whitespace changes
Inline Side-by-side
Showing with 335 additions and 0 deletions
docs/diagrams/agentic_hints_workflow.d2 0 → 100644
View file @ 218cb13c
# Render with reduced gap: d2 --layout=elk --elk-nodeNodeBetweenLayers=25 ...
direction: down
# Main box 1: Agentic client (title bold; subtitle plain)
agentic_client_box: "Agentic client" {
style.font-size: 28
shape: rectangle
direction: down
agentic_subtitle: "like NeMo Agent Toolkit" {
shape: text
style.font-size: 20
}
}
# Main box 2: Dynamo (big encapsulating box)
dynamo_box: "Dynamo" {
style.font-size: 28
direction: down
frontend_row: "" {
direction: right
frontend: "Dynamo Frontend" {
shape: rectangle
style.font-size: 24
}
frontend_parse: "Parses nvext → agentic_hints, cache_control" {
shape: text
style.font-size: 20
}
}
# One box per component (no wrapper around KV Router and KV Block Manager)
router_col: "" {
direction: down
kv_router: "KV Router" {
shape: rectangle
style.font-size: 24
}
kv_router_text: "- Latency sensitive first\n- Priority scheduling\n- Worker selection" {
shape: text
style.font-size: 20
}
}
blockmgr_col: "" {
direction: down
kv_block_mgr: "KV Block Manager" {
shape: rectangle
style.font-size: 24
}
kv_block_mgr_text: "- Cache Pinning\n- Prefetching\n- Eviction" {
shape: text
style.font-size: 20
}
}
}
# Footer: below Dynamo box, outside it
footer: "Every request can also include predictive metadata for 'expected future load', through estimates of number of requests, OSL and ITL." {
shape: text
style.font-size: 18
near: bottom-center
}
# Request label in top half of arrow (intermediate node to avoid overlap with Dynamo box)
request_label: "POST /v1/chat/completions\nbody: { ... \"nvext\": { \"agent_hints\": {...}, \"cache_control\": {...} } }" {
shape: text
style.font-size: 20
}
# Client -> label (top half) -> Frontend
agentic_client_box -> request_label
request_label -> dynamo_box.frontend_row.frontend
# Frontend -> KV Router and KV Block Manager
dynamo_box.frontend_row.frontend -> dynamo_box.router_col.kv_router
dynamo_box.frontend_row.frontend -> dynamo_box.blockmgr_col.kv_block_mgr
docs/features/agentic_workloads.md 0 → 100644
View file @ 218cb13c
---
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
title: Agentic Workloads
subtitle: Workload-aware inference with agentic hints for routing, scheduling, and KV cache Management
---
### Gaps with workload-agnostic inference
Agentic LLM inference is [dominated by KV-cache storage and I/O](https://arxiv.org/abs/2602.21548) rather than computation; without leveraging the predictable structure of agent lifecycles, we leave significant optimizations on the table.
Three gaps stand out with current workflows:
1. **Reactive vs. proactive:** Current runtimes do not use signals from the harness about what will happen next—e.g. that a "Plan" step is done and "Execute" steps are coming—so they cannot prefetch, pin, or schedule proactively.
2. **All KV-cache blocks treated equally:** Generic eviction (e.g. LRU) does not distinguish high-value, long-lived context (system prompt, tool definitions) from ephemeral context (chain-of-thought, scratchpad).
3. **Workload-agnostic scheduling:** Agents have predictable structure. Tools and system prompts repeat across turns, shallow vs. deep research have different latency needs, and the orchestrator knows which phase comes next.
## Dynamo as an Agentic Runtime
Dynamo exposes **agentic hints** and uses them at three layers: frontend API, router, and KV cache management. Together, these enable workload-aware inference instead of generic, state-of-the-moment optimization.
### Agentic Hints
Agentic hints are per-request metadata that the agent client (e.g. Claude Code, Codex, [NeMo Agent Toolkit](https://github.com/NVIDIA/NeMo-Agent-Toolkit)) sends to Dynamo's frontend. They are carried in the request body under [**nvext**](../components/frontend/nvext.md#agent-hints) on chat completions. The frontend parses them and passes them to the KV router and, where applicable, to the KV cache manager and backends.
- **Flow:** Harness sets hints in the request → Dynamo frontend parses `nvext` into routing hints → KV router uses them for queue ordering and worker selection → backends use them for priority scheduling and cache eviction.
![Agentic workflow: Harness → hints in request → Dynamo frontend → routing hints → KV router (queue order, worker choice) → backend](/assets/img/agentic-hints-workflow.svg)
The request body includes `nvext.agent_hints` (routing, scheduling) and `nvext.cache_control` (TTL-based pinning); the frontend passes the former to the KV router and the latter to the KV block manager for cache pinning, prefetching, and eviction.
| Hint | Description |
|------|-------------|
| `latency_sensitivity` | Router queue priority (requires `--router-queue-threshold`). Higher values shift the request earlier in the queue so user-facing turns run before background work. |
| `priority` | Engine queue ordering and KV cache eviction. Forwarded to the backend for scheduling and priority-based eviction. |
| `osl` | Expected output sequence length (tokens). Used by the router for output block tracking and load-balancing accuracy when `--router-track-output-blocks` is enabled. |
| `speculative_prefill` | When true, after the assistant turn completes the system prefills the predicted next-turn prefix (conversation history + assistant text, e.g. thinking stripped) to warm the KV cache for the next request. |
| `program_id` | (Planned) Identifies the agentic program for program-level metrics and cache affinity. |
| `context_type` | (Planned) Semantic type (e.g. system prompt, tool definition, reasoning branch) for context-aware eviction. |
**`nvext.cache_control`** (sibling of `agent_hints`, not inside it) provides TTL-based KV cache pinning. Pinned prefixes resist eviction for the specified duration. See [SGLang for Agentic Workloads — Cache Pinning](../backends/sglang/agents.md#cache-pinning-experimental).
## Feature matrix
| Feature | vLLM | SGLang | TensorRT-LLM |
|---------|:----:|:------:|:-------------:|
| Priority-based cache eviction | 🚧 | ✅ | 🚧 |
| Cache pinning | | ✅ | 🚧 |
| Cache prefetching | | 🚧 | |
| Subagent / thinking-aware cache eviction | | 🚧 | |
| Speculative prefill | ✅ | ✅ | ✅ |
| Latency-sensitivity–aware routing | ✅ | ✅ | ✅ |
🚧 = Work in progress or experimental.
### Using Dynamo from LangChain
Dynamo is now supported directly in LangChain using the [NVIDIA AI Endpoints integration](https://docs.langchain.com/oss/python/integrations/chat/nvidia_ai_endpoints#use-with-nvidia-dynamo). Configure the chat model to use the Dynamo endpoint and pass agent hints directly from the LangChain client.
## Features (experimental)
### KV cache optimizations
- **Priority-based KV cache eviction:** Instead of evicting by LRU alone, the backend can evict **low-priority** cache entries first when the GPU (and, with HiCache, host) cache is full. The `priority` value in `nvext.agent_hints` is forwarded to the engine; with SGLang, enable `--enable-priority-scheduling` and `--radix-eviction-policy priority`.
- **Cache pinning (experimental):** [Anthropic's v1/messages](https://docs.anthropic.com/en/docs/build-with-claude/caching) includes a `cache_control` field that tells servers how long to keep KV cache for specific blocks. Dynamo implements an OSS version with SGLang's HiCache: users can set `cache_control` via the same API as Anthropic or as an `nvext` field on chat completions. When set, the Dynamo router calls a hook in HiCache after the request completes to **pin** the blocks created by those tokens for the user-specified TTL. Pinned nodes resist eviction (demoting to host memory rather than being deleted).
In the Nemo Agentic toolkit and Dynamo integration, TTL is dynamically computed as the product of how many times a block is expected to be reused and the time between those requests; the NAT profiler pre-computes these expectations during agent evaluations and stores them in a data structure per agent, then injects `nvext.cache_control` with the derived TTL (see [dynamo_llm.py](https://github.com/NVIDIA/NeMo-Agent-Toolkit/blob/develop/packages/nvidia_nat_core/src/nat/llm/dynamo_llm.py)).
**Future work:** TTL could be determined dynamically by context type—e.g. think tokens or scratchpad content could use a lower TTL than system prompt or tool definitions, so high-value static context is retained longer while ephemeral context expires sooner.
- **Cache prefetching (future work):** Using the predictable agentic lifecycle (e.g. parent-child subagents, known next turn), Dynamo could proactively prefetch or move KV cache to a different worker so that the next request hits warm cache.
### Speculative prefill
After a turn finishes, the system can send a **speculative** `max_tokens=1` prefill with the **predicted next-turn prefix** (conversation history + assistant text, e.g. thinking stripped) to the same worker. When the real next request arrives, it hits a warm KV cache. Per-turn TTFT on turns 2+ can drop significantly (e.g. up to ~3× in [multiturn benchmarks](https://github.com/ai-dynamo/dynamo/blob/main/lib/bench/src/bin/README.md)). This can be extended so that Dynamo automatically sends tools and system prompt for subagents to a worker in advance, so subagent requests always hit warm cache.
### Latency-sensitivity–aware routing
When `--router-queue-threshold` is set, the router maintains a priority queue. Requests with higher `latency_sensitivity` are treated as if they arrived earlier, so they are scheduled ahead of bulk or background work. Under load, this keeps median latency low for user-facing agent turns while background work can tolerate higher latency. For a runnable demo and results, see [NeMo Agent Toolkit latency sensitivity demo](https://github.com/NVIDIA/NeMo-Agent-Toolkit/tree/develop/examples/dynamo_integration/latency_sensitivity_demo).
---
## See also
- [NeMo Agent Toolkit — Dynamo integration](https://github.com/NVIDIA/NeMo-Agent-Toolkit/tree/develop/examples/dynamo_integration)
- [Context engineering for AI agents (Manus)](https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus)
- [Stateful runtime for agents (OpenAI/Bedrock)](https://openai.com/index/introducing-the-stateful-runtime-environment-for-agents-in-amazon-bedrock/)
- [Claude Code's Prompt Caching](https://platform.claude.com/docs/en/build-with-claude/prompt-caching)
\ No newline at end of file
docs/index.yml
View file @ 218cb13c
......@@ -110,6 +110,11 @@ navigation:
path: agents/tool-calling.md
- page: LoRA Adapters
path: features/lora/README.md
- section: Agentic Workloads
path: features/agentic_workloads.md
contents:
- page: SGLang for Agentic Workloads
path: backends/sglang/agents.md
- section: Observability (Local)
path: observability/README.md
contents:
......
fern/assets/img/agentic-hints-workflow.svg 0 → 100644
View file @ 218cb13c
<?xml version="1.0" encoding="utf-8"?><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" data-d2-version="0.7.1" preserveAspectRatio="xMinYMin meet" viewBox="0 0 1148 899"><svg class="d2-959170537 d2-svg" width="1148" height="899" viewBox="-29 -29 1148 899"><rect x="-29.000000" y="-29.000000" width="1148.000000" height="899.000000" rx="0.000000" fill="transparent" class=" fill-N7" stroke-width="0" /><style type="text/css"><![CDATA[
.d2-959170537 .text {
font-family: "d2-959170537-font-regular";
}
@font-face {
font-family: d2-959170537-font-regular;
src: url("data:application/font-woff;base64,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");
}
.d2-959170537 .text-bold {
font-family: "d2-959170537-font-bold";
}
@font-face {
font-family: d2-959170537-font-bold;
src: url("data:application/font-woff;base64,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");
}
@media (prefers-color-scheme: dark) {
.d2-959170537 .fill-N1 { fill: #F2F2F2; }
.d2-959170537 .fill-N2 { fill: #E0E0E0; }
.d2-959170537 .fill-N3 { fill: #B0B0B0; }
.d2-959170537 .fill-N4 { fill: #404040; }
.d2-959170537 .fill-N5 { fill: #303030; }
.d2-959170537 .fill-N6 { fill: #252525; }
.d2-959170537 .fill-N7 { fill: transparent; }
.d2-959170537 .fill-B1 { fill: #F2F2F2; }
.d2-959170537 .fill-B2 { fill: #6B9FFF; }
.d2-959170537 .fill-B3 { fill: #5AA4DC; }
.d2-959170537 .fill-B4 { fill: #303040; }
.d2-959170537 .fill-B5 { fill: #252535; }
.d2-959170537 .fill-B6 { fill: #1E1E2E; }
.d2-959170537 .stroke-N1 { stroke: #F2F2F2; }
.d2-959170537 .stroke-N2 { stroke: #E0E0E0; }
.d2-959170537 .stroke-N3 { stroke: #B0B0B0; }
.d2-959170537 .stroke-N4 { stroke: #505050; }
.d2-959170537 .stroke-N5 { stroke: #404040; }
.d2-959170537 .stroke-N6 { stroke: #303030; }
.d2-959170537 .stroke-B1 { stroke: #F2F2F2; }
.d2-959170537 .stroke-B2 { stroke: #6B9FFF; }
.d2-959170537 .stroke-B3 { stroke: #5AA4DC; }
.d2-959170537 .color-N1 { color: #F2F2F2; }
.d2-959170537 .color-N2 { color: #E0E0E0; }
.d2-959170537 .color-N3 { color: #B0B0B0; }
.d2-959170537 .connection.fill-B1 { fill: #F2F2F2; }
}
]]></style><style type="text/css"><![CDATA[.shape {
shape-rendering: geometricPrecision;
stroke-linejoin: round;
}
.connection {
stroke-linecap: round;
stroke-linejoin: round;
}
.blend {
mix-blend-mode: multiply;
opacity: 0.5;
}
.d2-959170537 .fill-N1{fill:#0A0F25;}
.d2-959170537 .fill-N2{fill:#676C7E;}
.d2-959170537 .fill-N3{fill:#9499AB;}
.d2-959170537 .fill-N4{fill:#CFD2DD;}
.d2-959170537 .fill-N5{fill:#DEE1EB;}
.d2-959170537 .fill-N6{fill:#EEF1F8;}
.d2-959170537 .fill-N7{fill:#FFFFFF;}
.d2-959170537 .fill-B1{fill:#0D32B2;}
.d2-959170537 .fill-B2{fill:#0D32B2;}
.d2-959170537 .fill-B3{fill:#E3E9FD;}
.d2-959170537 .fill-B4{fill:#E3E9FD;}
.d2-959170537 .fill-B5{fill:#EDF0FD;}
.d2-959170537 .fill-B6{fill:#F7F8FE;}
.d2-959170537 .fill-AA2{fill:#4A6FF3;}
.d2-959170537 .fill-AA4{fill:#EDF0FD;}
.d2-959170537 .fill-AA5{fill:#F7F8FE;}
.d2-959170537 .fill-AB4{fill:#EDF0FD;}
.d2-959170537 .fill-AB5{fill:#F7F8FE;}
.d2-959170537 .stroke-N1{stroke:#0A0F25;}
.d2-959170537 .stroke-N2{stroke:#676C7E;}
.d2-959170537 .stroke-N3{stroke:#9499AB;}
.d2-959170537 .stroke-N4{stroke:#CFD2DD;}
.d2-959170537 .stroke-N5{stroke:#DEE1EB;}
.d2-959170537 .stroke-N6{stroke:#EEF1F8;}
.d2-959170537 .stroke-N7{stroke:#FFFFFF;}
.d2-959170537 .stroke-B1{stroke:#0D32B2;}
.d2-959170537 .stroke-B2{stroke:#0D32B2;}
.d2-959170537 .stroke-B3{stroke:#E3E9FD;}
.d2-959170537 .stroke-B4{stroke:#E3E9FD;}
.d2-959170537 .stroke-B5{stroke:#EDF0FD;}
.d2-959170537 .stroke-B6{stroke:#F7F8FE;}
.d2-959170537 .stroke-AA2{stroke:#4A6FF3;}
.d2-959170537 .stroke-AA4{stroke:#EDF0FD;}
.d2-959170537 .stroke-AA5{stroke:#F7F8FE;}
.d2-959170537 .stroke-AB4{stroke:#EDF0FD;}
.d2-959170537 .stroke-AB5{stroke:#F7F8FE;}
.d2-959170537 .background-color-N1{background-color:#0A0F25;}
.d2-959170537 .background-color-N2{background-color:#676C7E;}
.d2-959170537 .background-color-N3{background-color:#9499AB;}
.d2-959170537 .background-color-N4{background-color:#CFD2DD;}
.d2-959170537 .background-color-N5{background-color:#DEE1EB;}
.d2-959170537 .background-color-N6{background-color:#EEF1F8;}
.d2-959170537 .background-color-N7{background-color:#FFFFFF;}
.d2-959170537 .background-color-B1{background-color:#0D32B2;}
.d2-959170537 .background-color-B2{background-color:#0D32B2;}
.d2-959170537 .background-color-B3{background-color:#E3E9FD;}
.d2-959170537 .background-color-B4{background-color:#E3E9FD;}
.d2-959170537 .background-color-B5{background-color:#EDF0FD;}
.d2-959170537 .background-color-B6{background-color:#F7F8FE;}
.d2-959170537 .background-color-AA2{background-color:#4A6FF3;}
.d2-959170537 .background-color-AA4{background-color:#EDF0FD;}
.d2-959170537 .background-color-AA5{background-color:#F7F8FE;}
.d2-959170537 .background-color-AB4{background-color:#EDF0FD;}
.d2-959170537 .background-color-AB5{background-color:#F7F8FE;}
.d2-959170537 .color-N1{color:#0A0F25;}
.d2-959170537 .color-N2{color:#676C7E;}
.d2-959170537 .color-N3{color:#9499AB;}
.d2-959170537 .color-N4{color:#CFD2DD;}
.d2-959170537 .color-N5{color:#DEE1EB;}
.d2-959170537 .color-N6{color:#EEF1F8;}
.d2-959170537 .color-N7{color:#FFFFFF;}
.d2-959170537 .color-B1{color:#0D32B2;}
.d2-959170537 .color-B2{color:#0D32B2;}
.d2-959170537 .color-B3{color:#E3E9FD;}
.d2-959170537 .color-B4{color:#E3E9FD;}
.d2-959170537 .color-B5{color:#EDF0FD;}
.d2-959170537 .color-B6{color:#F7F8FE;}
.d2-959170537 .color-AA2{color:#4A6FF3;}
.d2-959170537 .color-AA4{color:#EDF0FD;}
.d2-959170537 .color-AA5{color:#F7F8FE;}
.d2-959170537 .color-AB4{color:#EDF0FD;}
.d2-959170537 .color-AB5{color:#F7F8FE;}.appendix text.text{fill:#0A0F25}.md{--color-fg-default:#0A0F25;--color-fg-muted:#676C7E;--color-fg-subtle:#9499AB;--color-canvas-default:#FFFFFF;--color-canvas-subtle:#EEF1F8;--color-border-default:#0D32B2;--color-border-muted:#0D32B2;--color-neutral-muted:#EEF1F8;--color-accent-fg:#0D32B2;--color-accent-emphasis:#0D32B2;--color-attention-subtle:#676C7E;--color-danger-fg:red;}.sketch-overlay-B1{fill:url(#streaks-darker-d2-959170537);mix-blend-mode:lighten}.sketch-overlay-B2{fill:url(#streaks-darker-d2-959170537);mix-blend-mode:lighten}.sketch-overlay-B3{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-B4{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-B5{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-B6{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-AA2{fill:url(#streaks-dark-d2-959170537);mix-blend-mode:overlay}.sketch-overlay-AA4{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-AA5{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-AB4{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-AB5{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-N1{fill:url(#streaks-darker-d2-959170537);mix-blend-mode:lighten}.sketch-overlay-N2{fill:url(#streaks-dark-d2-959170537);mix-blend-mode:overlay}.sketch-overlay-N3{fill:url(#streaks-normal-d2-959170537);mix-blend-mode:color-burn}.sketch-overlay-N4{fill:url(#streaks-normal-d2-959170537);mix-blend-mode:color-burn}.sketch-overlay-N5{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-N6{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.sketch-overlay-N7{fill:url(#streaks-bright-d2-959170537);mix-blend-mode:darken}.light-code{display: block}.dark-code{display: none}
@media (prefers-color-scheme: dark) {
.d2-959170537 .fill-N1 { fill: #F2F2F2; }
.d2-959170537 .fill-N2 { fill: #E0E0E0; }
.d2-959170537 .fill-N3 { fill: #B0B0B0; }
.d2-959170537 .fill-N4 { fill: #404040; }
.d2-959170537 .fill-N5 { fill: #303030; }
.d2-959170537 .fill-N6 { fill: #252525; }
.d2-959170537 .fill-N7 { fill: transparent; }
.d2-959170537 .fill-B1 { fill: #F2F2F2; }
.d2-959170537 .fill-B2 { fill: #6B9FFF; }
.d2-959170537 .fill-B3 { fill: #5AA4DC; }
.d2-959170537 .fill-B4 { fill: #303040; }
.d2-959170537 .fill-B5 { fill: #252535; }
.d2-959170537 .fill-B6 { fill: #1E1E2E; }
.d2-959170537 .stroke-N1 { stroke: #F2F2F2; }
.d2-959170537 .stroke-N2 { stroke: #E0E0E0; }
.d2-959170537 .stroke-N3 { stroke: #B0B0B0; }
.d2-959170537 .stroke-N4 { stroke: #505050; }
.d2-959170537 .stroke-N5 { stroke: #404040; }
.d2-959170537 .stroke-N6 { stroke: #303030; }
.d2-959170537 .stroke-B1 { stroke: #F2F2F2; }
.d2-959170537 .stroke-B2 { stroke: #6B9FFF; }
.d2-959170537 .stroke-B3 { stroke: #5AA4DC; }
.d2-959170537 .color-N1 { color: #F2F2F2; }
.d2-959170537 .color-N2 { color: #E0E0E0; }
.d2-959170537 .color-N3 { color: #B0B0B0; }
.d2-959170537 .connection.fill-B1 { fill: #F2F2F2; }
}
]]></style><g class="YWdlbnRpY19jbGllbnRfYm94"><g class="shape" ><rect x="300.000000" y="12.000000" width="297.000000" height="126.000000" stroke="#0D32B2" fill="#E3E9FD" class=" stroke-B1 fill-B4" style="stroke-width:2;" /></g><text x="448.500000" y="45.000000" fill="#0A0F25" class="text fill-N1" style="text-anchor:middle;font-size:28px">Agentic client</text></g><g class="ZHluYW1vX2JveA=="><g class="shape" ><rect x="12.000000" y="239.000000" width="1066.000000" height="547.000000" stroke="#0D32B2" fill="#E3E9FD" class=" stroke-B1 fill-B4" style="stroke-width:2;" /></g><text x="545.000000" y="272.000000" fill="#0A0F25" class="text fill-N1" style="text-anchor:middle;font-size:28px">Dynamo</text></g><g class="Zm9vdGVy"><g class="shape" ></g><text x="545.000000" y="824.000000" fill="#0A0F25" class="text fill-N1" style="text-anchor:middle;font-size:18px">Every request can also include predictive metadata for &#39;expected future load&#39;, through estimates of number of requests, OSL and ITL.</text></g><g class="cmVxdWVzdF9sYWJlbA=="><g class="shape" ></g><text x="449.000000" y="183.000000" fill="#0A0F25" class="text fill-N1" style="text-anchor:middle;font-size:20px"><tspan x="449.000000" dy="0.000000">POST /v1/chat/completions</tspan><tspan x="449.000000" dy="23.000000">body: { ... &#34;nvext&#34;: { &#34;agent_hints&#34;: {...}, &#34;cache_control&#34;: {...} } }</tspan></text></g><g class="YWdlbnRpY19jbGllbnRfYm94LmFnZW50aWNfc3VidGl0bGU="><g class="shape" ></g><text x="448.500000" y="82.000000" fill="#0A0F25" class="text fill-N1" style="text-anchor:middle;font-size:20px">like NeMo Agent Toolkit</text></g><g class="ZHluYW1vX2JveC5mcm9udGVuZF9yb3c="><g class="shape" ><rect x="281.000000" y="294.000000" width="723.000000" height="176.000000" stroke="#0D32B2" fill="#EDF0FD" class=" stroke-B1 fill-B5" style="stroke-width:2;" /></g></g><g class="ZHluYW1vX2JveC5yb3V0ZXJfY29s"><g class="shape" ><rect x="62.000000" y="560.000000" width="462.000000" height="176.000000" stroke="#0D32B2" fill="#EDF0FD" class=" stroke-B1 fill-B5" style="stroke-width:2;" /></g></g><g class="ZHluYW1vX2JveC5ibG9ja21ncl9jb2w="><g class="shape" ><rect x="544.000000" y="560.000000" width="484.000000" height="176.000000" stroke="#0D32B2" fill="#EDF0FD" class=" stroke-B1 fill-B5" style="stroke-width:2;" /></g></g><g class="ZHluYW1vX2JveC5mcm9udGVuZF9yb3cuZnJvbnRlbmQ="><g class="shape" ><rect x="331.000000" y="344.000000" width="235.000000" height="76.000000" stroke="#0D32B2" fill="#F7F8FE" class=" stroke-B1 fill-B6" style="stroke-width:2;" /></g><text x="448.500000" y="390.500000" fill="#0A0F25" class="text-bold fill-N1" style="text-anchor:middle;font-size:24px">Dynamo Frontend</text></g><g class="ZHluYW1vX2JveC5mcm9udGVuZF9yb3cuZnJvbnRlbmRfcGFyc2U="><g class="shape" ></g><text x="770.000000" y="389.000000" fill="#0A0F25" class="text fill-N1" style="text-anchor:middle;font-size:20px">Parses nvext → agentic_hints, cache_control</text></g><g class="ZHluYW1vX2JveC5yb3V0ZXJfY29sLmt2X3JvdXRlcg=="><g class="shape" ><rect x="112.000000" y="610.000000" width="151.000000" height="76.000000" stroke="#0D32B2" fill="#F7F8FE" class=" stroke-B1 fill-B6" style="stroke-width:2;" /></g><text x="187.500000" y="656.500000" fill="#0A0F25" class="text-bold fill-N1" style="text-anchor:middle;font-size:24px">KV Router</text></g><g class="ZHluYW1vX2JveC5yb3V0ZXJfY29sLmt2X3JvdXRlcl90ZXh0"><g class="shape" ></g><text x="378.500000" y="635.000000" fill="#0A0F25" class="text fill-N1" style="text-anchor:middle;font-size:20px"><tspan x="378.500000" dy="0.000000">- Latency sensitive first</tspan><tspan x="378.500000" dy="22.000000">- Priority scheduling</tspan><tspan x="378.500000" dy="22.000000">- Worker selection</tspan></text></g><g class="ZHluYW1vX2JveC5ibG9ja21ncl9jb2wua3ZfYmxvY2tfbWdy"><g class="shape" ><rect x="594.000000" y="610.000000" width="234.000000" height="76.000000" stroke="#0D32B2" fill="#F7F8FE" class=" stroke-B1 fill-B6" style="stroke-width:2;" /></g><text x="711.000000" y="656.500000" fill="#0A0F25" class="text-bold fill-N1" style="text-anchor:middle;font-size:24px">KV Block Manager</text></g><g class="ZHluYW1vX2JveC5ibG9ja21ncl9jb2wua3ZfYmxvY2tfbWdyX3RleHQ="><g class="shape" ></g><text x="913.000000" y="635.000000" fill="#0A0F25" class="text fill-N1" style="text-anchor:middle;font-size:20px"><tspan x="913.000000" dy="0.000000">- Cache Pinning</tspan><tspan x="913.000000" dy="22.000000">- Prefetching</tspan><tspan x="913.000000" dy="22.000000">- Eviction</tspan></text></g><g class="KGFnZW50aWNfY2xpZW50X2JveCAtJmd0OyByZXF1ZXN0X2xhYmVsKVswXQ=="><marker id="mk-d2-959170537-3488378134" markerWidth="10.000000" markerHeight="12.000000" refX="7.000000" refY="6.000000" viewBox="0.000000 0.000000 10.000000 12.000000" orient="auto" markerUnits="userSpaceOnUse"> <polygon points="0.000000,0.000000 10.000000,6.000000 0.000000,12.000000" fill="#0D32B2" class="connection fill-B1" stroke-width="2" /> </marker><path d="M 449.250000 140.000000 L 449.250000 159.000000" stroke="#0D32B2" fill="none" class="connection stroke-B1" style="stroke-width:2;" marker-end="url(#mk-d2-959170537-3488378134)" mask="url(#d2-959170537)" /></g><g class="KHJlcXVlc3RfbGFiZWwgLSZndDsgZHluYW1vX2JveC5mcm9udGVuZF9yb3cuZnJvbnRlbmQpWzBd"><path d="M 449.250000 211.000000 L 449.250000 340.000000" stroke="#0D32B2" fill="none" class="connection stroke-B1" style="stroke-width:2;" marker-end="url(#mk-d2-959170537-3488378134)" mask="url(#d2-959170537)" /></g><g class="ZHluYW1vX2JveC4oZnJvbnRlbmRfcm93LmZyb250ZW5kIC0mZ3Q7IHJvdXRlcl9jb2wua3Zfcm91dGVyKVswXQ=="><path d="M 410.083008 422.000000 L 410.083008 505.000000 S 410.083008 515.000000 400.083008 515.000000 L 197.498993 515.000000 S 187.498993 515.000000 187.498993 525.000000 L 187.498993 606.000000" stroke="#0D32B2" fill="none" class="connection stroke-B1" style="stroke-width:2;" marker-end="url(#mk-d2-959170537-3488378134)" mask="url(#d2-959170537)" /></g><g class="ZHluYW1vX2JveC4oZnJvbnRlbmRfcm93LmZyb250ZW5kIC0mZ3Q7IGJsb2NrbWdyX2NvbC5rdl9ibG9ja19tZ3IpWzBd"><path d="M 488.415985 422.000000 L 488.415985 505.000000 S 488.415985 515.000000 498.415985 515.000000 L 701.000000 515.000000 S 711.000000 515.000000 711.000000 525.000000 L 711.000000 606.000000" stroke="#0D32B2" fill="none" class="connection stroke-B1" style="stroke-width:2;" marker-end="url(#mk-d2-959170537-3488378134)" mask="url(#d2-959170537)" /></g><mask id="d2-959170537" maskUnits="userSpaceOnUse" x="-29" y="-29" width="1148" height="899">
<rect x="-29" y="-29" width="1148" height="899" fill="white"></rect>
</mask></svg></svg>
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment