Python SDK¶
The Topolograph Python SDK is a Pythonic, object-oriented client for the
REST API — plus a built-in SSH collector that gathers LSDBs from your devices,
and a topo CLI built on top of it.
topolograph-sdk on PyPI vadims06/topolograph-sdk
Install¶
Connect¶
from topolograph import Topolograph
topo = Topolograph(
url="http://localhost:8080",
token="your-api-token", # or set TOPOLOGRAPH_TOKEN
)
graph = topo.graphs.get(latest=True)
print(graph.graph_time, graph.protocol, graph.hosts['count'])
print(graph.status()['status'])
Authentication (in priority order)
- Token parameter —
Topolograph(url=..., token=...) - Environment —
export TOPOLOGRAPH_TOKEN=... - Basic auth —
Topolograph(url=..., username=..., password=...)
Collect topology over SSH¶
The SDK can log in to your devices, run the right per-vendor LSDB commands, and hand you the raw text — ready to upload.
from topolograph import TopologyCollector
collector = TopologyCollector("inventory.yaml")
result = collector.collect()
graph = topo.uploader.upload_raw(
lsdb_text=result.raw_lsdb_text,
vendor="FRR",
protocol="isis",
)
Inventory format¶
router1:
hostname: 172.20.20.2
username: admin
password: admin
vendor: frr
protocol: isis
port: 22
router2:
hostname: 172.20.20.3
username: admin
password: admin
vendor: cisco
protocol: ospf
Required per host: hostname, username, password, vendor
(cisco, juniper, frr, arista, nokia, huawei), and protocol
(ospf / isis). port is optional (defaults to 22). A starter file,
inventory.yaml.example, ships with the project.
Work with graphs¶
graphs = topo.graphs.list(protocol="ospf")
graph = topo.graphs.get_by_time("2024-01-15T10:30:00Z")
for node in graph.nodes.get():
print(node.name, node.id)
graph.networks.find_by_ip("10.10.10.1")
graph.networks.find_by_node("1.1.1.1")
graph.networks.find_by_network("10.10.10.0/24")
Compute paths¶
# Shortest path between nodes
path = graph.paths.shortest("1.1.1.1", "2.2.2.2")
print(path.cost)
for hops in path.paths:
print(" -> ".join(hops))
# Between IPs/networks
graph.paths.shortest_network("192.168.1.1", "192.168.2.1")
# Backup path (remove an edge and recompute)
graph.paths.shortest("1.1.1.1", "2.2.2.2",
removed_edges=[("1.1.1.1", "3.3.3.3")])
Read events¶
net = graph.events.get_network_events(last_minutes=60)
for e in net['network_up_down_events']:
print(e.event_object, e.event_status)
adj = graph.events.get_adjacency_events(
start_time="2024-01-15T10:00:00Z",
end_time="2024-01-15T11:00:00Z",
)
Filter links by TE attributes¶
graph.edges_list(temetric__gte=100)
graph.edges_list(unreserved_bw_0__lt=1e9)
graph.edges_list(src_node="1.1.1.1", dst_node="2.2.2.2", max_link_bw__gt=1e10)
See Traffic Engineering for the attribute list and operators.
MPLS TE tunnels¶
Read the CSPF placement result of the tunnels declared on a graph:
graph.lsps_list() # every tunnel path
graph.lsps_list(status="unplaced") # only what failed to place
graph.lsps_list(via_node="10.10.10.2") # paths crossing a node
graph.lsps_list(via_edge="10.10.10.1,10.10.10.2") # paths crossing a link
graph.lsps_list(include_path=True) # add the expanded node path
graph.lsp("TUN_R1_R3") # one tunnel, path always included
Each path carries placed, cost, and — when it failed — reason,
reason_code and binding_constraints. reason_code separates the two
failures that need opposite fixes: disconnected means no path exists even
with every constraint lifted (repair the topology), while
constraints_unsatisfiable means a path exists but the request is too strict
— relax the constraints named in binding_constraints (bandwidth,
affinity, srlg). Several entries mean they only block in combination.
Use via_edge_key instead of via_edge to pin an exact parallel/ECMP link;
get the key from graph.edges_list(include=["edge_key"]).
Managing tunnels:
graph.add_lsp({"name": "TUN_R1_R3", "src": "10.10.10.1", "dst": "10.10.10.3",
"bandwidth": "2G"})
graph.update_lsp("TUN_R1_R3", bandwidth="5G")
graph.delete_lsp("TUN_R1_R3")
graph.delete_lsps() # all tunnels on the graph
Check whether a path satisfying the constraints exists, without creating a tunnel — the check accounts for bandwidth already held by placed tunnels:
graph.cspf_path("10.10.10.1", "10.10.10.7",
bandwidth="5G",
metric_type="te",
admin_exclude_any=["red"],
srlg_exclude=[1001],
setup_priority=0)
# {'path': [...], 'cost': 42, 'reason': ''}
A plain shortest path ignores tunnels, as real IP forwarding does without
autoroute. Pass with_lsps=True to route over autoroute tunnels:
How much TE bandwidth a link has left once every placed tunnel is accounted for:
See MPLS TE Tunnels for the YAML key reference and the CSPF placement rules.
The topo CLI¶
The SDK installs a topo command:
# Graphs
topo graphs --list
topo graphs --latest
topo graphs --list --protocol ospf --watcher production-watcher
# Collect & upload
topo ingest inventory.yaml --protocol isis
topo ingest inventory.yaml --output lsdb.txt
topo ingest inventory.yaml --upload --url http://localhost:8080
# Paths
topo path --src 1.1.1.1 --dst 2.2.2.2
topo path --src 192.168.1.1 --dst 192.168.2.1 --network
topo path --src 1.1.1.1 --dst 2.2.2.2 --graph-time "2024-01-15T10:30:00Z"
# Upload an existing LSDB file
topo upload --file lsdb.txt --vendor FRR --protocol isis
topo upload --file lsdb.txt --vendor Cisco --protocol ospf --watcher prod-watcher
Error handling¶
from topolograph.exceptions import (
AuthenticationError, NotFoundError, ValidationError, APIError,
)
try:
graph = topo.graphs.get_by_time("invalid-time")
except NotFoundError:
...
except AuthenticationError:
...
except APIError as e:
print(f"API error: {e}")
Related: Getting topology in · MCP Server