# Task 2 Report — Reference-bundle ID evidence schema

## Status

DONE_WITH_CONCERNS

## Root cause

`_fasta_header_index()` preserved the complete FASTA header but extracted only
`gene` and `protein_id` metadata. Consequently, numeric NCBI GeneID evidence in
`[db_xref=GeneID:<digits>]` was not represented as a structured record field,
and `_build_reference_bundle_id_map()` had no `ncbi_gene_id` TSV column.

## Changes

- `apps/web/pipeline_backend.py` (reference-bundle parsing/schema)
  - Parse the exact numeric annotation form `[db_xref=GeneID:<digits>]` into a
    separate `ncbi_gene_id` record field.
  - Add `ncbi_gene_id` to `pep_cds_idmap.tsv`, using the existing CDS-first,
    PEP-fallback precedence.
  - Leave the existing `gene_id` extraction and selection unchanged, so
    `LOC110806180` remains the `gene_id` value.
  - Emit an empty string when neither FASTA record supplies numeric GeneID.
- `tests/unit/test_pipeline_backend_execution_env.py` (focused regression)
  - Add a real parser/schema test whose CDS header contains both
    `[gene=LOC110806180]` and `[db_xref=GeneID:110806180]`.
  - Assert the TSV retains `gene_id=LOC110806180` and separately emits
    `ncbi_gene_id=110806180`.
- No manifest count was added because Task 2 does not require one and the
  parser/schema behavior is sufficient.

## TDD evidence

### RED

Command:

```bash
/home/xia-lab/miniconda3/envs/test4py311/bin/python -m pytest tests/unit/test_pipeline_backend_execution_env.py -k reference_bundle_id_map -q
```

Result before production edits: exit code `1`; `1 failed, 30 deselected`.
The expected assertion path failed with:

```text
KeyError: 'ncbi_gene_id'
```

The preceding `gene_id == "LOC110806180"` assertion passed, confirming the
failure was specifically the missing numeric GeneID schema field.

### GREEN — focused regression

The same command after the minimal production edit exited `0` with:

```text
1 passed, 30 deselected in 0.08s
```

### GREEN — required focused suite

Command:

```bash
/home/xia-lab/miniconda3/envs/test4py311/bin/python -m pytest tests/unit/test_pipeline_backend_execution_env.py -k 'reference_bundle or ppi' -q
```

Result: exit code `0`; `25 passed, 6 deselected in 1.83s`.

### Syntax compilation

Command:

```bash
/home/xia-lab/miniconda3/envs/test4py311/bin/python -m py_compile apps/web/pipeline_backend.py
```

Result: exit code `0`; no output.

## Scope review

The complete diffs against `/tmp/ppi-id-resolution-task2-baseline/` contain
only:

- the two parser record lines and two TSV schema/row lines in
  `_fasta_header_index()` / `_build_reference_bundle_id_map()`;
- one `csv` test import and the focused regression test;
- this report.

PPI querying, Ortholog logic, frontend behavior, canonical IDs, contracts,
manifests, and unrelated stages were not changed. No service or Celery restart
was performed because Task 2 validation directly imports the current module and
does not exercise a running service. No diagnostic record was updated because
the failure is covered by the new regression test and does not add runtime
diagnostic behavior. No real-data/API execution was run because the accepted
Task 2 brief explicitly scopes validation to the local reference-bundle parser
and focused unit suite.

## Concerns

- Pytest emitted the pre-existing `RequestsDependencyWarning` that installed
  `urllib3`/character-detection package versions do not match the version range
  expected by `requests`. It did not cause test failures, but the environment
  dependency skew remains outside Task 2 scope.