Repository navigation
Releases
v0.5.2
v0.5.1
Fixed
- ADIF is read by
qso_graph_lib.adif, the fleet's one ADI reader, which reads ADI as
ADIF 3.1.7 defines it and passes ADIF's own test files.qsp-adif-convertand
qsp_client.adif.readerhad a scanner each, and they lost data:reader.pylooked for tags anywhere, values included, so a COMMENT containing
<CALL:5>W9XYZreplaced the QSO's callsign.- Both read files in text mode, which turns CR LF into LF and breaks every multi-line
field's declared length;qsp-adif-convertalso dropped undecodable bytes
(errors="ignore"), shifting every length after them. Files are now read with their
line ends and bytes as written. - A file with no Header no longer has its first record's fields taken as the Header's.
The functions and the converter's output are otherwise unchanged. New dependency:
qso-graph-lib>=0.2.0(standard library only).
v0.5.0
Changed
-
Python 3.11 is the minimum. 3.10 reached end of life in October 2026
(PEP 619), so it gets no further security
releases on any platform. On Windows it was worse than unsupported:uv's 3.10
is built against OpenSSL 1.1.1w, end of life since September 2023, so HTTPS
from this server ran over an unmaintained TLS library.certifidoes not help
with that — it supplies the root store, not the implementation.3.10 support was already nominal. This package has linted with
ruff target-version = "py311"and type-checked withmypy python_version = "3.11"for some time, so nothing ever verified the 3.10 claim.uv.lockis regenerated and the diff is large, but nothing is upgraded —
it is the 3.10 resolution branch disappearing. Packages that carried two pins,
one per Python branch, now carry the single 3.11+ pin they already resolved to.
An install on 3.11 or later is unchanged.
Fixed
ci-all-greencould not fail on the dependency audit. Theauditjob was
added to itsneedslist on 2026-10-26 without the matching result check, and
withif: always()theneedslist gates nothing — the check script does. A
lockfile advisory would have turned the audit red and leftci-all-greengreen:
a gate that cannot fail, in the repository where two advisories had just been
found. The result is now checked, as in the other 23.- The
http://check exempted10.and192.168.unanchored, so they matched
anywhere in a line andhttp://example.com/api/v10.2/thingpassed — verified
with a planted file. Neither exemption protected anything:src/holds exactly
onehttp://URL, the localhost default inconfig.py. The check is now the
fleet's form,localhost/127.0.0.1/::1only, and prints the offending line
rather than only reporting that one exists. An exemption names the exact thing,
never a prefix (.claude/rules/security.md). - The workflow actions move off Node 20:
actions/checkout@v4→@v5,
actions/setup-python@v5→@v6,astral-sh/setup-uv@v6→@v7. The other 23
repositories moved on 2026-10-26; this one was missed because it sits outside the
fleet's repository selector — which is also how its missing Release page and its
retired release flow went unnoticed.
v0.4.1
Security
certifiinuv.lockmoved from 2026.7.22 to 2026.10.23, clearing a CRITICAL
advisory (GHSA-6q48-6c7w-9gw2). Nobody installing from PyPI was exposed:
certifiarrives transitively throughhttpxwith an open range anduv.lock
is not in the wheel, so a fresh install always resolved a currentcertifi.
What was exposed is anything running from this repository with
uv sync --frozen, which includes this repository's own CI.opentelemetry-apimoved from 1.45.0 to 1.45.1, clearing CVE-2026-114603.
Found by the new audit job below, not by Dependabot, which had not reported it:
pip-auditreads the lockfile against an advisory database, and that is a
different question from the one Dependabot's alert list answers.
Changed
- A release is now a deliberate act, not a side effect of merging.
publish.ymlfires onworkflow_dispatchwith the version typed in, and
merging intomainpublishes nothing. This was the last repository still on
thedevelop→main→ publish-on-merge flow that the 22 MCP servers left on
2026-10-25; itsdevelopwas five commits behindmain, so that branch had
already stopped being where work landed.developis gone. - CI audits the locked dependencies, per pull request and weekly. Its absence is
why thecertifiadvisory above sat unnoticed: nothing readuv.lockagainst
an advisory database. The weekly run matters because an advisory is published
against a lockfile that has not changed, and no push or pull request fires then. - The one-off
deprecate-qsp-mcpjob is removed. The MCP Registry already
reportsio.github.qso-graph/qsp-mcp0.1.1 asstatus=deprecated, so it is
spent.
v0.4.0
- ADIF tools moved here from
adif-mcp, which is now only the MCP server and ADIF's published
resources.qsp-adif-convertconverts an ADIF (.adi) log to JSON or NDJSON (it was
adif-mcp convert);qsp_client.adif.readerreads.aditext andqsp_client.adif.models
holds the QSO and entity models. They came with their tests. Nothing in the relay uses them. qsp-adif-convertnever drops a record silently (#24). A record it cannot convert used to
vanish unless--errorswas given, and the run still exited 0, so a log with no station
callsign converted to[]and looked like success. It now always reports how many records
failed and why on stderr, says to use--station-callwhen that is the reason, and exits 1.
v0.3.5
- ruff and mypy run in CI (qso-graph-devel#66), as a job the
ci-all-greengate requires. This
repo sat outside that rollout; it is in the MCP ecosystem and answers to the same rules as the
servers. Settings followadif-mcp, as everywhere else, and both run once rather than per Python
version: they read the source, and neither answer changes with the interpreter. _discover_toolsbuilds rows of(server, namespaced name, tool), which is what
mcp_to_openai_tools(namespace=True)documents and reads, but the list was annotated as holding
two-tuples. Nothing produced or consumed that shape; the annotation was simply wrong.BridgeConfigwas imported intorelay.pyand never used, and two functions promised adict
and astrwhile returning whatever the JSON body held.E501is deferred rather than adopted (qso-graph-devel#70): what it reports in these repos are
widths, not defects.
v0.3.4
- The published contact is
maintainers@qso-graph.io(qso-graph-devel#69). Theauthorsfield
carried a personal address, and that field is what PyPI shows on the package page. The project
has had outside contributions; a project address is the fitting route for them.
v0.3.3
v0.3.2
- PyPI: the Documentation link goes to this package's own page, https://qso-graph.io/servers/qsp-client/ (qso-graph/.github#15).
- CI: the release flow (qso-graph/.github TEMPLATES.md). Work lands on
develop; a release is a
PR fromdevelopintomain, and merging it publishes to PyPI and the MCP Registry, verifies both
and tags the release. CI runs ondeveloptoo, and PRs intomainmust come fromdevelopor a
security/branch.
v0.3.1
Fixed
- Tool names are always valid function names (#8). When servers share a tool name (every
qso-graph server hasget_version_info), the copies are namespaced asserver__tool, which
0.3.0 already did. The server part comes from your config, though, so a name likemy solaror
9975.wsprcould produce a function name that OpenAI or Gemini rejects, and so could a tool's
own name (get version,solar.wind), which comes from a third-party server. Every exposed
name is now cleaned to[A-Za-z0-9_-], starts with a letter or underscore, fits in 64
characters, and stays unique even when two names clean to the same text; each still calls its
server under its original name. Tests cover the namespacing, shared and unshared names, and the
64-character limit with suffixes.