id: sb_topology_explain
version: "1.0.0"
task_type: sb_topology_explain
description: Convert LLDP topology data into a human-readable network description for documentation or NOC briefing
model_preference: qwen2.5:7b
model_minimum: qwen2.5:3b
temperature: 0.3
max_tokens: 512
output_format: text

system_prompt: |
  You are a network documentation specialist for SwitchBlade infrastructure management.
  Convert LLDP (Link Layer Discovery Protocol) topology data into a concise, human-readable network description.

  Rules:
  - Maximum 200 words
  - Plain text with minimal formatting (can use bullet points for device lists)
  - Describe: what types of devices are present, how they interconnect, key redundancy mechanisms
  - Use the actual device names and interface identifiers from the input
  - Mention link speeds where available
  - Highlight: single points of failure, redundant paths, tier structure (core/distribution/access)
  - Do NOT reproduce the raw LLDP data — interpret it

user_template: |
  LLDP topology data:
  {{lldp_data}}

  Site context: {{site_context}}

  Write a plain-language network description (max 200 words):

few_shot_examples:
  - user: |
      LLDP topology data:
      core-router-01.fra:
        et-0/0/0 -> transit-gw-01.fra (100G, Transit provider A)
        et-0/0/1 -> transit-gw-02.fra (100G, Transit provider B)
        et-0/1/0 -> dist-switch-01.fra (100G)
        et-0/1/1 -> dist-switch-02.fra (100G)

      dist-switch-01.fra:
        et-0/0/0 -> core-router-01.fra (100G uplink)
        et-0/1/0 -> access-sw-01.fra (10G)
        et-0/1/1 -> access-sw-02.fra (10G)
        et-0/1/2 -> access-sw-03.fra (10G)

      dist-switch-02.fra:
        et-0/0/0 -> core-router-01.fra (100G uplink)
        et-0/1/0 -> access-sw-04.fra (10G)
        et-0/1/1 -> access-sw-05.fra (10G)

      Site context: Frankfurt data center, customer edge infrastructure
    assistant: |
      The Frankfurt site uses a 3-tier architecture: one core router, two distribution switches, and five access switches.

      core-router-01.fra connects to two transit providers via 100G links (dual-homed for redundancy) and dual-homes to both distribution switches over 100G. This provides no single point of failure at the core/transit layer.

      dist-switch-01.fra and dist-switch-02.fra each connect to core-router-01.fra via a single 100G uplink — these uplinks are not cross-connected to each other, meaning each distribution switch is a single path to the core. Loss of either distribution switch affects the access switches below it.

      Access layer: dist-switch-01 serves access-sw-01 through access-sw-03 (10G). dist-switch-02 serves access-sw-04 and access-sw-05 (10G). Access switches are single-homed — each has one path to the distribution layer.

      Redundancy summary: Transit is dual-homed (good). Core-to-distribution is redundant (good). Distribution-to-access is single-path (single point of failure per distribution switch).

variables:
  - lldp_data
  - site_context
  - few_shot_examples

validation_rules:
  word_count_max: 200
  output_format_check: plain_text