Files
gns3-server/docs/gns3-copilot/implemented/fault-injection.md
T
YueGuobin 6a4cbe39c1 refactor(copilot): reuse MCP tools via a bridge; extract shared device runners
The copilot and the MCP service maintained two parallel wrapper layers
(LangChain BaseTool classes vs @mcp.tool() functions) over the same
handlers in gns3_client/api_handlers.py. Descriptions and parameters
drifted apart (compute_id and link_type were missing on the copilot
side). This makes the MCP registration the single source of truth and
removes the duplicated wrappers.

tools_v2/mcp_bridge.py adapts the MCP server's tool registration into
LangChain tools: names, descriptions and argument schemas come from the
MCP tools. The pool is filtered to project-scoped tools
(MCP_ALLOWED_TOOLS) because the copilot runs inside one open project —
project management, compute/appliance/image administration and server
info tools stay out. The bridge carries the copilot's request JWT into
the MCP contextvars (username/token_version decoded from the token so
ticket-minting tools work) and runs the async MCP functions from the
copilot's sync tool node on either thread flavor.

Mode toolsets (teaching / lab_automation / troubleshooting_injection)
are now composed from adapted MCP tools plus a few copilot-native ones
that add value the MCP versions lack: link_create (port-name
resolution), wait_seconds, packet_analysis(+skills), and the topology
reader. Eight orphaned LangChain wrappers are deleted.

The device command tools had the dependency inverted (the MCP handlers
instantiated the copilot LangChain classes and round-tripped through
JSON strings). The Nornir/Netmiko classes moved verbatim (git mv) into
the shared layer as plain runners — DeviceConfigCommandRunner,
DeviceShowCommandRunner, VpcsCommandRunner — with _run(json_string)
becoming run(params, gns3_ctx). The optional Jinja2 template rendering
moved from the MCP handlers into the runners (and into
utils/device_configs.render_device_template), so both entry paths get
it; the MCP handlers are now pure delegation. This also drops the
json.dumps/json.loads detour the MCP path used to take.

test_tool_params consistency gate extended: a handler that forwards the
whole params dict to a callee is a wildcard (same rationale as the
params.items() case) — the per-key check cannot see through delegation.
2026-11-03 22:04:57 +08:00

5.0 KiB

This documentation is organized by AI with reference to actual code. AI can make mistakes — please verify against the source code when in doubt.

Fault Injection Feature

Overview

The fault injection feature enables GNS3 Copilot to automatically inject realistic network faults into GNS3 labs for troubleshooting training. The agent analyzes the topology, selects an appropriate fault based on the configured protocols, injects it, and documents the results.

Architecture

graph TD
    subgraph "Agent Workflow"
        TG[Topology Gathering]
        CFG[Config Collection]
        FA[Fault Analysis]
        FI[Fault Injection]
        FD[Fault Documentation]
    end

    subgraph "Skills Repository"
        IS[INJECTION_SKILLS_REGISTRY<br/>39 categories, 368 issues]
        SK[injection_skill_get Tool]
    end

    subgraph "Tools"
        DC[device_show_run]
        CC[device_config_send]
        GK[gns3_topology_reader]
    end

    TG -->|Get topology| GK
    CFG -->|Get configs| DC
    FA -->|Query filtered skills| IS
    FA -->|Select fault| SK
    FI -->|Inject config changes| CC
    FD -->|Output report| FD

Injection Skills Tool

The injection_skill_get (LangChain BaseTool) is the primary interface for querying available faults.

Listing Faults (with context filter)

The LLM MUST pass topology context when listing faults:

{"action": "list", "context": ["ospf", "bgp", "vlan"]}

This returns only faults matching the protocols found in the topology. The tool rejects calls without context:

{
  "error": "context parameter is required when action='list'",
  "hint": "Analyze the topology and device configurations first...",
  "available_categories": ["bgp", "interface", "mpls", "ospf", ...]
}

Getting Fault Details

Token-efficient usage pattern:

// Step 1: List issue names only (~300 tokens)
{"device_type": "injection_ospf_issues", "detail": "index"}

// Step 2: Get single issue detail (~500 tokens)
{"device_type": "injection_ospf_issues", "issue": "ospf_hello_dead_mismatch"}

Parameters

Parameter Required Description
action No ("get") "list" to browse, "get" for details
context Yes for action="list" Protocols found in topology: ["ospf", "bgp"]
device_type Yes for action="get" e.g. "injection_ospf_issues"
detail No ("full") "index" (names), "summary" (+desc), "full" (all)
issue No Single issue key for targeted detail

Fault Injection Workflow

sequenceDiagram
    participant LLM
    participant injection_skill_get
    participant TopologyTool
    participant DeviceCommands

    LLM->>TopologyTool: Get topology info
    TopologyTool-->>LLM: Topology + device types

    LLM->>DeviceCommands: Get device configs
    DeviceCommands-->>LLM: Running configurations

    Note over LLM: Analyze which protocols are in use

    LLM->>injection_skill_get: {"action": "list", "context": ["ospf", "vlan"]}
    injection_skill_get-->>LLM: Matching fault categories

    LLM->>injection_skill_get: {"device_type": "injection_ospf_issues", "detail": "index"}
    injection_skill_get-->>LLM: Issue names (token-efficient)

    LLM->>injection_skill_get: {"device_type": "injection_ospf_issues", "issue": "ospf_hello_dead_mismatch"}
    injection_skill_get-->>LLM: Full fault detail + config commands

    LLM->>DeviceCommands: Inject fault configuration
    DeviceCommands-->>LLM: Execution result

    Note over LLM: Document fault in response

Injection Skills Repository

Skills are organized by protocol/category in the external GNS3-Skills repository:

Category File Example Issues
OSPF injection/ospf_issues.yaml Hello/Dead mismatch, MTU mismatch, area mismatch
BGP injection/bgp_issues.yaml AS-path prepend, next-hop unreachable, route filtering
VLAN injection/vlan_issues.yaml Trunk allowed mismatch, native VLAN mismatch
STP injection/stp_issues.yaml Root guard, loop guard, port priority
MPLS injection/mpls_issues.yaml LDP session down, label binding failure
... 34 more files 368 issues total

Recovery

Each injected fault includes restore commands in the documentation. The LLM always provides commands to fully revert all changes.

API Endpoint

POST /copilot/projects/{project_id}/chat/inject

Dedicated endpoint for fault injection. Internally sets copilot_mode to troubleshooting_injection and runs the agent.

Request:

{
  "message": "Inject a network fault for troubleshooting practice",
  "session_id": "optional-session-uuid"
}

Response: Server-Sent Events (SSE) stream identical to the chat stream endpoint.