Practical self-hosting · intermediate
20 · Every model needs a passport
Track provenance, configuration, rights and runtime compatibility.
What you can take away
- Create a versioned model manifest with explicit unknowns.
- Choose a small model you are already permitted to inspect, or use the fictional manifest
Your trail through this idea
Checking saved progress…
Ian asked about reusing downloaded models across LocalAI-related services. A file being present on disk is only the beginning: format, runtime, configuration, and permitted use still need to match.
Prerequisite: complete the preceding lesson or bring its worksheet.
Understand the system
Ollama’s Modelfile reference shows how model construction and runtime parameters can be described. A model manifest for your service should also record source, immutable identifier where available, format, quantization, intended task, license location, tested runtime, context setting, and evaluation results. A friendly alias is not enough to recreate a tested configuration.
Licenses and model cards are model-specific. The availability of weights does not establish permission for every intended use. This module supplies a recordkeeping method, not a universal license interpretation or assurance that any model fits your hardware.
Work through the case
Original manifest fields:
name: DEMO_MODEL
source: OFFICIAL_MODEL_PAGE
revision: UNVERIFIED
format: UNVERIFIED
runtime: TEST_RUNTIME_AND_VERSION
license: REVIEW_REQUIRED
context: MEASURED_SETTING
status: not admitted
An unknown field is a gate, not an invitation to guess.
Make something useful
Try this
Original exercise
Choose a small model you are already permitted to inspect, or use the fictional manifest. Find its official model card and runtime documentation. Complete only fields you can support. Write one capability test and one failure test. Do not download large weights or replace a serving model for this exercise. Store any local paths privately.
Check your reasoning
A quick check · for your own learning
What should an unknown license or runtime requirement do in a model manifest?
A self-study prompt. You can retry; this is not an exam.
Keep the result and follow the sources
Success is a manifest another operator can evaluate and reproduce. Model aliases may change; the manifest preserves what was actually tested. Promotion should wait until required unknowns are resolved.
Sources and scope
- Primary documentation or archival source, reviewed 14 November 2026. Additional direct links appear beside the claims they support.
- Editorial basis: Ian's historical self-hosting questions and current repository inspection informed the case selection. Historical requests do not establish current deployment. Private chat records, credentials and host details are excluded.
- Examples and exercises are original educational synthesis. External sources support the stated technical or historical context, not every hypothetical scenario.