feat(schema): complete hardware tables with GPU identity, lookups, and slot links

[AlbinoGeek]

Implements the schema asked for in #38 (and #17–#22), for @gissf1's review before it lands, per the approval requirement on #38.

Design choices where the issues were unsettled (newer issue wins):

1. vram_type is an FK to a seeded vram_types table (per Tables are not defined properly and missing many fields. #38 and the Hardware tables disk persistence — analysis and layout #22 comment).
2. Per-format TFLOPS columns dropped in favour of gpu_tflops (GPU × compute_formats × tflops_sources).
3. PCI/TDP names follow Tables are not defined properly and missing many fields. #38: pci_subsystem_device_id, tdp_w.
4. lane_transfer_rate is a float (PCIe 1.0 is 2.5 GT/s).
5. PCIe bandwidth uses published bidirectional figures (PCIe 4.0 x16 = 63.02 GB/s), per schema: implement interface_type lookup table with seed data #17's acceptance criteria.
6. SXM/NVLink use per-GPU aggregate bandwidth (300/600/900 GB/s); Thunderbolt 10 GB/s; OCuLink carries PCIe 4.0 x4 values.
7. USB4, eGPU and integrated seed rows dropped (not in schema: implement interface_type lookup table with seed data #17); an integrated GPU has a null native interface.
8. New lookup tables use plural names, like model_architectures.
9. schema_version_id on system_hardware left out, pending the UUID-migration design.

Validated by loading all of schema.sql in PGlite: seed counts correct, duplicate PCI identity and bad slot links rejected. gate green. No application code reads these tables yet (auto-detection per #18–#20 is not started).

Still open: the run_hardware_metrics fields at the end of #8 that #21 didn't cover.

Refs #38

🤖 Generated with Claude Code

[Copilot]

Copilot review overview

🟡 Changes recommended

The unresolved critical GPU-history issue and two moderate compatibility/slot-link issues must be addressed.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1 · 2

Open (3)

• Mutable GPU slot links rewrite historical run hardware identity · New
• Replacing cpu_cores drops the collector's physical core count · New
• Allowing independently nullable hardware slot key columns · New

What changed in this PR

This pull request completes the hardware schema for GPU identity, lookup data, interconnects, TFLOPS provenance, and system-slot relationships.

Changes:

• Adds lookup seed data for VRAM, architectures, compute formats, TFLOPS sources, and interfaces.
• Expands hardware, GPU, system, slot-link, and run-metrics relationships.
• Documents the schema additions in the changelog.

File	Summary
schema/​seeds/​vram_types.json	Seeds VRAM technologies.
schema/​seeds/​tflops_sources.json	Seeds TFLOPS provenance sources.
schema/​seeds/​interface_types.json	Seeds PCIe and interconnect types.
schema/​seeds/​gpu_architectures.json	Seeds GPU architectures.
schema/​seeds/​compute_formats.json	Seeds compute formats.
schema/​schema.sql	Defines expanded hardware relationships. Findings: Moderate (3 votes): enforce paired nullability for slot-link identifiers. Critical (1 vote): preserve GPU identity for historical metrics. Moderate (2 votes): retain cpu_cores or update the collector and persistence contract.
CHANGELOG.md	Documents the schema changes.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.