Commit cfcf532
feat(julia-server): wire /gnn/rank to the trained solver
`api/gnn_endpoint.jl` was never included in EchidnaML nor registered
with the HTTP handler — every call to POST /gnn/rank hit the main
server's 404 branch, so the Rust GNN client always fell back through
`reqwest` to cosine similarity regardless of model availability.
Changes
- EchidnaML.jl: include api/gnn_endpoint.jl after api/server.jl so
GNN_MODEL / GNN_VOCAB globals and route handlers are resolvable.
- api/server.jl: after `init_server_state` loads the NeuralSolver,
publish it via `GNN_MODEL[]` and compose the HTTP handler with
`register_gnn_routes!` so /gnn/rank and /gnn/health are served.
Effect: once run_training.jl produces models/neural/final_model and
the server is started against that path, /gnn/rank ranks via the
trained PremiseRanker rather than the cosine fallback.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>1 parent cf581a3 commit cfcf532
2 files changed
Lines changed: 13 additions & 1 deletion
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
188 | 188 | | |
189 | 189 | | |
190 | 190 | | |
| 191 | + | |
191 | 192 | | |
192 | 193 | | |
193 | 194 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
265 | 265 | | |
266 | 266 | | |
267 | 267 | | |
| 268 | + | |
| 269 | + | |
| 270 | + | |
| 271 | + | |
| 272 | + | |
| 273 | + | |
| 274 | + | |
| 275 | + | |
268 | 276 | | |
269 | 277 | | |
270 | 278 | | |
271 | 279 | | |
272 | 280 | | |
273 | 281 | | |
| 282 | + | |
| 283 | + | |
274 | 284 | | |
275 | | - | |
| 285 | + | |
| 286 | + | |
276 | 287 | | |
277 | 288 | | |
278 | 289 | | |
| |||
0 commit comments