| name | netbox-diode |
| description | Patterns for ingesting data into NetBox using the Diode SDK (Python and Go). Use when building network discovery, bulk import, or automated data population integrations that write data to NetBox via gRPC. Covers SDK setup, entity construction, batch ingestion, chunking, reconciler behavior, error handling, and when to use Diode vs the REST API.
|
| license | Apache-2.0 |
NetBox Diode SDK
Diode is a gRPC-based data ingestion service for NetBox. Instead of managing dependency order and object IDs via the REST API, you describe objects by name and the Diode reconciler resolves dependencies, creates missing objects, and performs upserts automatically.
This skill covers the open-source Python and Go SDKs (diode-sdk-python v1.12.0 / diode-sdk-go v1.9.0), targeting NetBox 4.5.x–4.6.x. The Diode server and reconciler are proprietary — this skill describes their observable behavior, not internals.
Your knowledge of Diode SDK may be outdated. Entity types, SDK methods, and reconciler behavior evolve between releases. Prefer retrieval over pre-trained knowledge.
Retrieval Sources
| Source | URL / Method | Use for |
|---|
| Diode docs | https://netboxlabs.com/docs/diode/ | Architecture, getting started |
| Diode Python SDK | https://netboxlabs.com/docs/diode/sdk/python/ | Python SDK reference |
| Python SDK repo | https://github.com/netboxlabs/diode-sdk-python | Source, entity types, changelog |
| Go SDK repo | https://github.com/netboxlabs/diode-sdk-go | Source, entity types, changelog |
| Protobuf reference | https://netboxlabs.com/docs/diode/diode-proto/ | Entity schema definitions |
| NetBox MCP server | If configured — verify objects created by Diode landed correctly | Post-ingestion validation |
For a quick comparison of Diode vs REST API, see
netbox-api-integration/references/diode-integration.md.
FIRST: Verify Setup
Confirm the Diode SDK is installed and your credentials are configured:
pip show netboxlabs-diode-sdk
echo $DIODE_CLIENT_ID
echo $DIODE_CLIENT_SECRET
If the SDK is not installed: pip install netboxlabs-diode-sdk
To verify Diode server connectivity:
from netboxlabs.diode.sdk import DiodeClient
with DiodeClient(target="grpc://YOUR_DIODE_HOST:8080/diode", app_name="test", app_version="0.1") as client:
print("Connected")
When to Use Diode
| Scenario | Recommended |
|---|
| Network discovery pushing many objects | Diode |
| Bulk import / migration | Diode |
| Creating many related objects (devices + interfaces + IPs) | Diode |
| Reading / querying NetBox data | REST or GraphQL (Diode is write-only) |
| Single object CRUD with immediate read-back | REST API |
| Complex filtered searches | REST or GraphQL |
Key advantages over REST: No dependency ordering, no ID lookups, automatic object creation, gRPC performance for high volume.
Key limitation: Write-only. You cannot query NetBox through Diode.
Quick Reference
Client Types
| Client | Python | Go | Purpose |
|---|
| Standard | DiodeClient | diode.NewClient | Production ingestion (requires Diode server) |
| Dry Run | DiodeDryRunClient | diode.NewDryRunClient | JSON output for testing (no server needed) |
| OTLP | DiodeOTLPClient | diode.NewOTLPClient | Export as OpenTelemetry log records |
Authentication
OAuth2 client credentials via environment variables (recommended):
export DIODE_CLIENT_ID="your-client-id"
export DIODE_CLIENT_SECRET="your-client-secret"
Or pass directly: client_id= / client_secret= (Python), WithClientID() / WithClientSecret() (Go).
Minimal Python Example
from netboxlabs.diode.sdk import DiodeClient
from netboxlabs.diode.sdk.ingester import Device, Interface, IPAddress, Entity
with DiodeClient(
target="grpc://localhost:8080/diode",
app_name="my-discovery",
app_version="1.0.0",
) as client:
entities = [
Entity(device=Device(
name="sw-01",
device_type="Catalyst 9300",
manufacturer="Cisco",
site="NYC-DC1",
role="Access Switch",
status="active",
)),
Entity(interface=Interface(device="sw-01", name="Gi0/1", type="1000base-t")),
Entity(ip_address=IPAddress(address="10.0.1.1/24")),
]
response = client.ingest(entities=entities)
if response.errors:
for err in response.errors:
print(f"Error: {err}")
Minimal Go Example
import diode "github.com/netboxlabs/diode-sdk-go/diode"
client, err := diode.NewClient(
"grpc://localhost:8080/diode",
"my-discovery",
"1.0.0",
diode.WithClientID("YOUR_CLIENT_ID"),
diode.WithClientSecret("YOUR_CLIENT_SECRET"),
)
if err != nil { log.Fatal(err) }
defer client.Close()
entities := []diode.Entity{
&diode.Device{
Name: diode.String("sw-01"),
DeviceType: &diode.DeviceType{Model: diode.String("Catalyst 9300")},
Site: &diode.Site{Name: diode.String("NYC-DC1")},
Role: &diode.DeviceRole{Name: diode.String("Access Switch")},
Status: diode.String("active"),
},
}
resp, err := client.Ingest(context.Background(), entities)
Entity Construction
String Shorthand (Python only)
Python entities with a primary key in PRIMARY_VALUE_MAP accept string shorthand for nested references:
Device(name="sw-01", site="NYC-DC1")
Device(name="sw-01", site=Site(name="NYC-DC1"))
Use full objects when you need to set attributes beyond the primary key:
Device(name="sw-01", site=Site(name="NYC-DC1", status="active", region="US-East"))
Go has no string shorthand. Always use full struct types with pointer helpers (diode.String(), diode.Int()).
Required Fields
Most entities only require their primary identifying field (e.g., name for Device, address for IPAddress). The reconciler creates referenced objects automatically. See references/entity-catalog.md for the full list.
Key Uniqueness Constraints
| Entity | Unique By |
|---|
| Device | name + site + tenant |
| Interface | name + device |
| IPAddress | address + vrf |
| Prefix | prefix + vrf |
| VLAN | vid + group |
Reconciler Behavior
The reconciler is the server-side component that processes ingested data. Key observable behaviors:
- Automatic dependency resolution — referencing
site="NYC-DC1" creates the Site if it doesn't exist
- Name-based matching — objects matched by their primary identifying field
- No ordering required — unlike REST, you don't need to create manufacturers before device types
- Upsert semantics — matching entities are updated, not duplicated
- Slug auto-generation —
slug derived from name if not provided
- Case-sensitive matching —
"Site ABC" ≠ "site abc"
- Producer provenance — entities tagged with
app_name / app_version for audit
Ingestion Patterns
Batch Ingestion
Send multiple entities in one call:
response = client.ingest(entities=[
Entity(device=device1),
Entity(device=device2),
Entity(interface=iface1),
], metadata={"batch_id": "scan-001"})
Chunked Ingestion (large batches)
gRPC has a ~4MB message limit. Use built-in chunking for large payloads:
from netboxlabs.diode.sdk import create_message_chunks
for chunk in create_message_chunks(entities):
client.ingest(entities=chunk)
Go: client.Ingest(ctx, entities, diode.WithChunking(0)) (0 = default 3MB).
Dry Run and OTLP
See references/ingestion-patterns.md for dry run testing, replay, and OTLP export.
Error Handling
Python Exception Hierarchy
DiodeConfigError # Bad credentials, unreachable auth endpoint
DiodeClientError # gRPC transport errors (inherits grpc.RpcError; has .code(), .details())
OTLPClientError # OTLP export failures
Note: DiodeClientError inherits from grpc.RpcError, so except grpc.RpcError also catches it.
Auto-Retry on Auth Expiry
The SDK automatically retries on UNAUTHENTICATED gRPC errors (token expiry), up to 3 times by default (max_auth_retries or DIODE_MAX_AUTH_RETRIES).
Recommended Pattern
from netboxlabs.diode.sdk.exceptions import DiodeClientError, DiodeConfigError
try:
with DiodeClient(target=target, app_name="app", app_version="1.0") as client:
response = client.ingest(entities=entities)
if response.errors:
for err in response.errors:
log.warning(f"Entity error: {err}")
except DiodeConfigError as e:
log.error(f"Config/auth error: {e}")
except DiodeClientError as e:
log.error(f"gRPC error: {e.status_code} - {e.details}")
Anti-Patterns
- IPAddress requires CIDR —
"192.168.1.1/24", not "192.168.1.1"
- Interface requires device — name alone isn't unique without device context
- Go has no string shorthand — every nested reference needs a full struct
- Diode is write-only — use REST/GraphQL to read data back
- Case-sensitive matching — inconsistent casing creates duplicate objects
- gRPC 4MB limit — use chunking for large batches
- String shorthand limitations — can't set nested object attributes (use full objects)
- Metadata must be JSON-compatible — strings, numbers, booleans, lists, dicts only
References
Cross-Skill References