refactor(cli)!: drop the generated Stainless command tree and the CLI generator - #2004
Draft
maxdubrinsky wants to merge 25 commits into
Draft
refactor(cli)!: drop the generated Stainless command tree and the CLI generator#2004maxdubrinsky wants to merge 25 commits into
maxdubrinsky wants to merge 25 commits into
Conversation
Adds what hand-written CLI commands need from nemo_platform_plugin: - InferenceGatewayClient (provider/model/openai proxy routes, provider_ready, raw SSE streams, OpenAI model listing) with endpoint and wire tests. - client_from_platform accepts a NemoClient or AsyncNemoClient and derives the typed client with from_client, so callers no longer need to know which platform handle they hold; PlatformClient is the runtime-checkable structural type for that parameter and platform_default_headers reads identity headers off either shape. - models.refs holds the pure model-reference helpers; packages/models re-exports them so the CLI does not import the Stainless-bound package. - filesets.transfer holds SDK-free upload/download/list/delete; FilesResource delegates to it and filesets no longer imports .resources eagerly. - Bearer tokens are resolved on every HTTP attempt rather than baked into the PreparedRequest, so retries and later pages never replay a stale token. - A 409 on a create sent with exist_ok is no longer retried before send() resolves it by fetching the existing entity. - Query-param TypedDicts and request models for files, iam, virtual models and workspaces gain the fields the CLI exposes; GuardrailConfig.data is optional to match the server entity. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
…ands Additive plumbing so command groups can be rewritten on the typed clients one PR at a time while the generated commands keep working: - client/bootstrap.py resolves config, OIDC discovery and token providers without the generated SDK and builds NemoClient/AsyncNemoClient with the same retry policy, TLS verification and 5 s connect cap the SDK used; factory.py layers the NeMoPlatform constructors on top of it. - CLIContext.typed_client(XClient) / async_typed_client derive a service client that shares the platform client's transport and auth; get_workspace exposes the configured default. - pagination gains collect_offset_pages / collect_cursor_pages for typed paginated responses, carrying the server's envelope fields (sort, filter, grouped_by) into JSON output; fetch_all_pages stays for generated commands. - errors maps typed-client and pydantic errors alongside the SDK ones (validation and unknown --input-data keys exit 2), and recognises every unresolved-workspace message. - stdin_utils.build_request_body validates --input-data into a request model and rejects unknown keys instead of dropping them. - code_generator renders typed-client snippets (request models as constructor calls, SecretStr masked, RootModel positional); generated commands are routed to legacy_code_generator until they are replaced. - waiters accept either platform handle, poll with a monotonic clock and normalise str-enum statuses; formatters unwrap NemoResponse; the version flag reads distribution metadata; command groups no longer advertise shell completion (the root app owns it). Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
…p calls it The new test endpoint stubs raise NotImplementedError like their siblings instead of an ellipsis body, which ty rejects as an implicit None return. The auth-idp CLI refresh contract test patches discover_nmp_config on client.bootstrap, where the CLI now resolves OIDC settings. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
AllPagesResponse and OffsetPageResponse emit the server envelope fields, but callers that build them without an envelope (fetch_all_pages behind the generated list commands, and nemo jobs list --all-pages) lost the sort key that list output has always carried. sort is now always present, null when the server did not echo one, and a real envelope keeps its own field order. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
- Narrow the platform handle from object to PlatformClient in client_from_platform and platform_default_headers (adapter.py) and the nemo.sdk resource-factory owner in NemoPluginSDKResources (sdk.py), so the typed-client boundary stops erasing the type where plugins are built. - build_request_body reports each field's accepted input alias in the unknown-input hint (schema, not the internal schema_) so the advertised key is one pydantic actually accepts. - with_options now clears a clone's cached nemo.sdk plugin resources, which were built against the original client's transport, so typed clients accessed on the clone bind to its headers/retry/timeout. - Drop the always-false _PLATFORM_JOB_LIFECYCLE watch exclusions from the legacy code generator's lifecycle helpers. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
Brings in the auth-off-Stainless SDK removal, customization typed-client migration, and sandbox forward merges. Resolves conflicts in adapter.py and client.py by combining both sides: keep main's _owns_http lifecycle (owns_http_client=False) and _platform_default_headers header resolution, and keep the PlatformClient narrowing plus the with_options cached-resource clearing added for review. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
This was referenced Sep 11, 2026
The main merge kept the typed resource factories on PlatformClient but took main's __all__ which no longer re-exports NeMoPlatform/AsyncNeMoPlatform, so the import became dead and tripped the ruff pre-commit hook (lint-python-style). Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
Plugins (nemo-evaluator among others) import the generated NeMoPlatform / AsyncNeMoPlatform classes from nemo_platform_plugin.sdk. My earlier drop of that import broke them with an ImportError. Re-export lazily through a module __getattr__ so plugin discovery still imports this module without requiring the generated SDK at load time. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
Brings in guardrails typed-client (#1966) and other main changes. Resolved guardrail/types.py to main's landed GuardrailConfig.data shape (RailsConfig, typed) per the layered-merge rule, and took main's add/add tests/guardrail/test_endpoints.py. Also fixes the sdk.py re-export regression in the same push. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
maxdubrinsky
force-pushed
the
cli-drop-stainless/mdubrinsky
branch
from
September 11, 2026 18:48
0273a21 to
884b00b
Compare
maxdubrinsky
changed the base branch from
cli-typed-client-plumbing/mdubrinsky
to
main
September 11, 2026 18:50
Contributor
Contributor
|
…bing-merge Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com> # Conflicts: # packages/nemo_platform_ext/src/nemo_platform_ext/cli/core/errors.py
…bing-merge Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
…bing-merge Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
…bing-merge Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com> # Conflicts: # packages/nemo_platform_plugin/tests/client/test_adapter.py
maxdubrinsky
force-pushed
the
cli-drop-stainless/mdubrinsky
branch
from
September 11, 2026 21:15
884b00b to
e44bfc5
Compare
…bing-merge Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
… clients The four groups (including the inference sub-groups: providers, models, prompts, deployments, deployment-configs, virtual-models, gateway) become hand-written modules on FilesClient, ModelsClient and InferenceGatewayClient via state.typed_client(), replacing their generated Stainless commands and the generator overrides that patched them. Flags, defaults, choices, table columns and help text are unchanged; the generator skips these resources. Fileset upload/download/list/delete go through filesets.transfer. files list -f code renders the same query the command sends and refuses a glob --remote-path it cannot express; nemo files otlp logs create sends real payloads (it previously sent an empty body) labelled application/json. --wait on deployments create polls through ModelsClient and the gateway readiness endpoint with a monotonic clock. Wire-level tests cover every command over a recorded transport; integration tests run against the in-process Files and Models services. `files filesets refresh` (added to the generated tree by #1957) is carried over: FilesClient.refresh_fileset posts to the /refresh route and the command renders the returned fileset, with the same help text and position. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
Hand-written Typer groups for workspaces (with members), projects and iam
on the typed clients from the plumbing PR, replacing their generated
Stainless commands. The CLI surface, output shapes and exit codes are
unchanged: list commands go through collect_offset_pages so --all-pages and
the {data, sort, pagination} envelope match the generated output, and
--input-data bodies go through build_request_body, which rejects unknown
fields with exit code 2 and the accepted field list.
The generated workspaces and projects modules are removed and the generator
config marks them skip: true so make generate-cli-commands cannot bring them
back. The generated iam group was already removed by the core auth port
(#1968); this adds the hand-written hidden group back on IAMClient.
Secrets is not part of this PR: #1971 landed its own port of that group.
Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
… packages The three functional groups move out of the generated command tree and become nemo.cli entry points on the packages that own the services: guardrail in nemo-guardrails-plugin (GuardrailCLI), intake and experiments in nmp-intake (IntakeCLI, ExperimentsCLI). They appear only when that package is installed, the same way every other plugin group does, and are written on GuardrailClient and IntakeClient via state.typed_client(). Command names, flags, defaults and columns are unchanged; the group help line becomes the standard "Plugin commands for <name>." row. The intake typed client gains the read and experiment endpoints the CLI needs. The nmp-intake bundle inherits nemo.* entry points so the nemo-platform wheel actually carries the groups (the generated entry-point table is regenerated with make vendor), and a new test asserts every bundled package's nemo.* entry points are exposed by the wrapper or SDK pyproject so a missing inherit fails in CI rather than in the shipped wheel. The generator skips the three resources. GuardrailConfig.data is Optional again after #1966 typed it as a bare RailsConfig: the guardrails entity stores it as Optional and the API returns "data": null for configs created without a body, which the CLI integration tests here create. The middleware already handled None. Rebased over #1965, which added its own experiment endpoints and read models: main's ExperimentCreateRequest/UpdateRequest/Response (dict-typed pareto and column_layout) and RetrieveTraceQueryParams are used; the endpoint stubs keep the create exist_ok / list / delete variants the CLI needs, and span groups go back through collect_offset_pages now that list_span_groups returns Paginated[SpanGroup] on main. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
… plugin runtime Plugin discovery, the scheduler, seed jobs, job result plumbing, the job API factory and the plugins that build on them (nemo-agents, nemo-optimization, nemo-agent-hardener, anonymizer, auditor, evaluator, safe-synthesizer, hello-world) annotated the platform handle they receive as the generated NeMoPlatform / AsyncNeMoPlatform classes even though they only forward it to client_from_platform. They now use the structural PlatformClient type, so the plugin runtime and nemo-agents import without the generated SDK installed. nemo_platform_plugin.sdk keeps re-exporting the generated classes for plugins that still import them from there. jobs.watch classifies transient failures by shape (transport error or retryable HTTP status) instead of by SDK exception class. chat_tui decodes SSE itself and consumes typed BinaryContent streams; the generated SDK's streaming context manager is adapted until nemo chat moves over. The agent-hardener CLI builds a NemoClient directly. Rebased over #1973, which reworked nemo-agents' call sites: main's versions of those files win. This PR keeps the three nemo-agents modules #1973 left on the generated SDK (optimize_cli, usage/cli, usage/sources/fileset) plus the other plugins. nemo-optimization resolves platform agents through AgentsClient instead of the generated sdk.agents surface, since that attribute does not exist on the typed client; the test serves the agent over a MockTransport instead of stubbing it. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
setup builds a frozen SetupClients bundle (ModelsClient, SecretsClient, InferenceGatewayClient) from the CLI context instead of walking the generated SDK's resource tree, so the wizard no longer depends on nemo_platform for provider, secret and gateway readiness calls. Behaviour is unchanged: the same probes, prompts, telemetry events and error paths. Adds wire-level tests for the setup calls over a recorded transport and retargets the onboarding telemetry tests at the typed bundle. The Kind e2e for auto model selection drove _run_auto_mode with a raw NeMoPlatform handle; it now builds SetupClients from the same handle with client_from_platform, as the rest of the e2e suite already does. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
…dubrinsky Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com> # Conflicts: # packages/nemo_platform_ext/src/nemo_platform_ext/cli/commands/api/__init__.py # packages/nemo_platform_ext/src/nemo_platform_ext/cli/commands/manifest_registry.py # packages/nemo_platform_ext/tests/cli/integration/conftest.py # tools/nemo-platform-sdk-tools/src/nemo_platform_sdk_tools/sdk/cli_generator/cli_config.yaml
…nless/mdubrinsky Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com> # Conflicts: # packages/nemo_platform_ext/src/nemo_platform_ext/cli/commands/api/__init__.py # packages/nemo_platform_ext/tests/cli/test_app.py # tools/nemo-platform-sdk-tools/src/nemo_platform_sdk_tools/sdk/cli_generator/cli_config.yaml
…nless/mdubrinsky Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
…less/mdubrinsky Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
… generator With every group hand-written on typed clients, the CLI no longer needs the generated nemo_platform SDK: - CLIContext.get_client() returns a NemoClient built by client.bootstrap (OAuth refresh, workload identity, API key and direct modes preserved); typed_client() derives service clients from it. - nemo chat, nemo wait, the agent context command and quickstart use typed clients; chat_tui consumes typed SSE streams only. - The generated cli/commands/api tree, its top-level entry table, the legacy code-generation and pagination paths that served it, and the Jinja CLI generator in nemo-platform-sdk-tools (with the generate-cli-commands make target and lint-fix step) are removed. A plugin nemo.cli entry point can still replace a same-named built-in group in the Core or Functional panels. - tests/cli/test_stainless_boundary.py enforces the boundary: a static scan of the CLI package plus a subprocess that makes nemo_platform un-importable and runs every core group, plugin discovery, -f code invocations and the plugin-hosted groups. - The nmp-cli skill and package docs describe the hand-written model. Based on the jobs port in #1922, which the rewritten jobs group relies on. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
route adapter import the generated SDK at module level. Two client-side modules reached them transitively: entity_client re-exported get_entity_client eagerly (pulling in dependencies), and the plugin CLI command builder imported two wire constants from functions.routes. get_entity_client is now resolved lazily through a module __getattr__, and the NDJSON media type and default function path live in functions.frames (shared by adapter and CLI) with routes re-exporting them. Plugin discovery, seed jobs and plugin CLIs import cleanly with nemo_platform un-importable, which the boundary test asserts with no remaining allow-list. The generated CLI generator tree is removed with the command tree. Signed-off-by: Max Dubrinsky <mdubrinsky@nvidia.com>
maxdubrinsky
force-pushed
the
cli-drop-stainless/mdubrinsky
branch
from
September 11, 2026 21:53
e44bfc5 to
193c740
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Final PR of the series:
CLIContext.get_client()returns a typedNemoClient, the generatedcli/commands/api/*tree and the CLI generator (tools/nemo-platform-sdk-tools/.../cli_generator,make generate-cli-commands,make update-cli) are deleted, and the boundary test asserts that every CLI group, plugin discovery, seed jobs and plugin CLIs import withnemo_platformmade un-importable. After this thenemoCLI has no dependency on the Stainless SDK.Depends on the middle PRs of the stack (core groups 1 and 2, plugin-hosted groups, #2002 plugin runtime typing, #2003 setup); this branch merges them, so its diff shrinks to the flip as they land. Marked
!becausenemo_platform_ext.cli.core.context.CLIContext.get_clientchanges type and the-f codeoutput for every command now emits typed-client Python instead ofNeMoPlatformcalls.Changes
cli/core/{context,errors,pagination,waiters,code_generator}.py,app.py,version: Stainless branches removed;errorsmaps pluginApiError/ValidationError/UnknownInputFieldsErroronly.cli/commands/api/and the generator tree deleted;Makefile,tools/lint/lint-fix.sh,tools/mcp-dev-tools/nmp_dev_mcp.py, sdk-toolscli.pylose the generate targets;legacy_code_generator.pyremoved.nemo_platform_plugin:sdk.pyandentity_client.pyresolve their Stainless-dependent re-exports lazily via module__getattr__, and the NDJSON wire constants move tofunctions/frames.py, so plugin discovery no longer imports the generated SDK transitively (refactor(sdk): route platform clients through owned endpoints #1920 added module-level SDK imports todependencies.pyandfunctions/routes.py, which are left as they are).tests/cli/test_stainless_boundary.py: no allow-list; probes the root app, every group's--helpand plugin discovery with aMetaPathFinderthat blocksnemo_platform*..agents/skills/nmp-cli/SKILL.md,packages/nemo_platform_ext/{AGENTS.md,README.md,docs/overview.md},nemo_platform_plugin/client/MIGRATION.mdrewritten for hand-written groups andnemo.clientry points.Stack
Part of the AIRCORE-893 series that moves the
nemoCLI off the generated Stainless SDK onto the typed clients innemo_platform_plugin(blocking AIRCORE-827). Stacked on #1986 (plumbing) but targeted atmainso CI runs (theCIworkflow only triggers for PRs againstmain/release/*). Until #1986 merges, the diff includes its commits; this PR's own change is the merge commits of #1999, #2000, #2001, #2002, #2003 plus the two commitsrefactor(cli)!: drop the generated Stainless command tree and the CLI generatorandfix(plugin): keep the generated SDK off the plugin discovery import path. Review that commit, or compare againstcli-typed-client-plumbing/mdubrinsky.get_client()'s Stainless path, deletecli/commands/api/*and the CLI generatorStatus: draft. Being maintained by an agent on behalf of @mdubrinsky; it re-bases on #1986 as that moves and on @ironcommit's landing plugin migrations (#1971, #1973), and picks up CI and review feedback. It will be re-cut where those PRs overlap (noted per PR below).
Type of Change
Quality Gates
Verification
Signed-off-by:traileruv run pre-commit run -apasses, or any blocked checks are identified belowTargeted validation (on the current head, based on #1986 at
377fb4d0a):make test-unit(whole repo): 18325 passed, 59 skipped, 2 xfailed; 13 failures are the known host port conflicts (test_services.py,nemo-deployments/test_ports.py,nemo-agents/test_port_allocation.py) and pass in CI.bash tools/lint/lint-python-types.sh: passes.make generate-cli-reference-docs: no diff beyond the three plugin-hosted help lines from the plugin-hosted groups PR.NMP_CONFIG_FILE=/tmp/nope.yaml uv run _nemo --helpand every group--helpwithnemo_platformblocked at import: pass (the boundary test).uv run ruff checkanduv run ruff format --check: clean.uv run pre-commit run -a: all hooks pass exceptuv-lock(host uv 0.9.30 vs pinned 0.9.14);uvx --from uv==0.9.14 uv lock --checkpasses.helm-docsandyqare not installed on the host; both hooks are CI-covered.NMP_CONFIG_FILE=/tmp/nope.yaml uv run _nemo <group> --helpdiffed againstmainfor every rewritten group: identical, except a trailing period on group help lines.