backend: give the reference library the MCP tools it was missing, and fix the one that broke

The reference library landed with a full REST API and no MCP surface at all:
an agent could cite `reference_id: 42` — the citation is validated now — but
could not look up what 42 was, create it, or find out whether the library
already had it. That is the citation half of "generate the paper, then write it
in" missing, and it is the half an agent is worst at doing by hand.

Twelve tools, mirroring the REST surface plus one addition each way:

- `reference_*` — search, read, create, update, delete, plus
  `reference_import`, which takes a whole pasted bibliography in one call and
  reports what it skipped (an existing DOI or an identical title) and what it
  could not read, one bad entry at a time rather than losing the other nineteen;
  and `reference_citations`, which answers "what is this cited for" before a
  rename or a delete. `reference_list`'s keyword search across title, container,
  DOI and author names is the de-duplication path that `reference_import` uses.
- `reference_tag_*` — the five CRUD tools. Deleting a tag drops the labelling
  and leaves the literature, which is the opposite of what the field library
  needs and the same choice the REST route makes.

Two things had to be fixed to get here, both found by testing rather than
reading:

`template_list` was crashing. The rename that turned a heading into a title and
its use into a placement changed `TemplateListItem.field_count` to
`placement_count` and left the tool reading the old attribute, so the call died
with `Error executing tool template_list` — unreadable, and live in the
cluster. The suite passed 47 checks while this was broken because it never
called `template_list`: the uncovered tools were exactly the read-only ones, so
the write paths were all green and the browse path was not. The suite now
covers all 41 tools — 76 checks — and step 5 asserts the shape each read tool
returns, not merely that it returned.

`DOMAIN_REFUSALS` in the tool registry held one class, `UnknownReference`. The
tag rules arrived with the tag tools and raise two more, so 「标签不存在:9」 and
「标签「必读」已存在」 would have reached a model as `Error executing tool …`.
The tuple is now the list of every refusal a tool is expected to translate, and
a new one belongs in it on the day it is written.

`reference_update` takes its changes as one nested object rather than 38 keyword
parameters: a nested model keeps `model_fields_set`, so an omitted field stays
absent while an explicit null clears the column. Flattened, the two are
indistinguishable — every omitted field arrives as None — which is the same
problem `paper_update` solves by rebuilding its payload from what it was given.

Verified against the deployed service, not a local process: 76 checks pass over
the NodePort, then the same 76 over stdio and against a local HTTP server, and
the REST suite still passes its 55. The image is rebuilt as v4 and the
Deployment follows it, because `imagePullPolicy: IfNotPresent` means re-pushing
a tag the node already holds changes nothing.
This commit is contained in:
2026-09-22 18:30:24 +08:00
parent a29badd362
commit 15955f552d
10 changed files with 1085 additions and 75 deletions
+15 -1
View File
@@ -464,6 +464,8 @@ written by an agent is indistinguishable from one written by hand. What
| `paper_write` / `paper_write_text` | "generate the paper, then put it in" is one intention, not thirty round trips |
| `paper_delete` refuses once before it deletes | a cascading delete has no undo in a tool call |
| `sentence_search` across papers | the job is consistency — one paper says 洪水损失, the next must not say GUL |
| `reference_*` and `reference_tag_*` | the reference library is what makes a citation mean something; an agent that can cite id 42 but cannot look up what 42 *is* is writing blind |
| `reference_import` | "here are the twenty references I cited" is one intention, and twenty round trips would spend twenty schemas on it |
Two transports are served from one build: `stdio`, which is what a client
spawns, and `streamable-http`, which is what a client on another machine
@@ -486,7 +488,19 @@ a database that is briefly unreachable cannot become a restart loop, while
readiness runs `SELECT 1` so the pod leaves the Service's endpoints while TiDB
is down and rejoins when it is not.
Two decisions are worth knowing before adding a tool. **Results are registered
Two shape decisions are worth knowing before adding a tool. **A nested argument
model keeps `model_fields_set`**, which a flattened signature cannot: 38
bibliographic fields spread across keyword parameters would arrive with the
omitted ones indistinguishable from explicit nulls, so `reference_update` takes
its changes as one object and forwards only the keys the caller actually wrote.
And **a domain refusal belongs in `DOMAIN_REFUSALS`** in `tools/registry.py` on
the day it is written — `UnknownReference` was translated there when the
reference tools were still to come, and the tag rules (`UnknownTag`,
`DuplicateTagName`) had to be added when those tools landed, because an
unlisted refusal reaches the model as `Error executing tool …` and says nothing
about what to fix.
Two more decisions are worth knowing. **Results are registered
with `structured_output=False`**: inferred from a `-> str` annotation, the SDK
publishes a `{"result": …}` envelope and sends the JSON twice, once as
`structuredContent` and once as text, and clients that read only one of the two