Skip to content

armgit / dify

Public Imported from GitHub
0

Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.

test(api): strictly type agent prompt and skill package tests (#45221)

Eddy ZHANG170dbaeb3478
diff --git a/api/tests/unit_tests/.ruff.toml b/api/tests/unit_tests/.ruff.toml
index f28b9056483..4f8e0503ffc 100644
--- a/api/tests/unit_tests/.ruff.toml
+++ b/api/tests/unit_tests/.ruff.toml
@@ -225,7 +225,6 @@ extend-select = ["ANN401", "ARG"]
"models/test_dataset.py" = ["ARG005"]
"services/agent/test_observability_service.py" = ["ARG002", "ARG005"]
"services/agent/test_roster_service.py" = ["ARG001", "ARG002", "ARG003", "ARG005"]
-"services/agent/test_prompt_mentions.py" = ["ARG005"]
"../helpers/unit/services/controller_api.py" = ["ARG002"]
"services/data_migration/test_import_service.py" = ["ARG002", "ARG005"]
"../helpers/unit/services/dataset_service_test_helpers.py" = ["TID251"]
diff --git a/api/tests/unit_tests/pyrefly.toml b/api/tests/unit_tests/pyrefly.toml
index 1d4e97eb67b..41c1d421c12 100644
--- a/api/tests/unit_tests/pyrefly.toml
+++ b/api/tests/unit_tests/pyrefly.toml
@@ -343,8 +343,6 @@ project-excludes = [
"repositories/test_sqlalchemy_api_workflow_run_repository.py",
"services/agent/test_dsl_service.py",
"services/agent/test_roster_service.py",
- "services/agent/test_prompt_mentions.py",
- "services/agent/test_skill_package_service.py",
"services/agent/test_workflow_publish_service.py",
"../helpers/unit/services/controller_api.py",
"services/data_migration/test_export_service.py",
diff --git a/api/tests/unit_tests/services/agent/test_prompt_mentions.py b/api/tests/unit_tests/services/agent/test_prompt_mentions.py
index 7bba5236f50..ef24da1442b 100644
--- a/api/tests/unit_tests/services/agent/test_prompt_mentions.py
+++ b/api/tests/unit_tests/services/agent/test_prompt_mentions.py
@@ -15,6 +15,7 @@ from services.agent.prompt_mentions import (
NODE_JOB_PROMPT_ALLOWED_KINDS,
SOUL_PROMPT_ALLOWED_KINDS,
MentionKind,
+ PromptMention,
build_node_job_mention_resolver,
build_soul_mention_resolver,
expand_prompt_mentions,
@@ -29,7 +30,7 @@ from services.agent.prompt_mentions import (
# ── parse ─────────────────────────────────────────────────────────────────────
-def test_parse_extracts_kind_id_and_optional_label():
+def test_parse_extracts_kind_id_and_optional_label() -> None:
prompt = "Use [§skill:tender-analyzer%2FSKILL.md:tender-analyzer§] then ask [§human:c-1§]."
mentions = parse_prompt_mentions(prompt)
@@ -40,25 +41,25 @@ def test_parse_extracts_kind_id_and_optional_label():
assert prompt[mentions[0].start : mentions[0].end] == mentions[0].raw
-def test_parse_supports_ids_with_slash_and_dot():
+def test_parse_supports_ids_with_slash_and_dot() -> None:
mentions = parse_prompt_mentions("[§tool:langgenius/tavily/tavily_search:tavily§] [§node_output:node-1.tenders§]")
assert mentions[0].ref_id == "langgenius/tavily/tavily_search"
assert mentions[1].ref_id == "node-1.tenders"
-def test_parse_supports_legacy_bare_output_tokens():
+def test_parse_supports_legacy_bare_output_tokens() -> None:
mentions = parse_prompt_mentions("Use §output:summary:summary§")
assert [(mention.kind, mention.ref_id, mention.label) for mention in mentions] == [
(MentionKind.OUTPUT, "summary", "summary")
]
-def test_parse_ignores_legacy_template_forms_and_unknown_kinds():
+def test_parse_ignores_legacy_template_forms_and_unknown_kinds() -> None:
prompt = "{{var}} {{#context#}} {{#sys.query#}} [§bogus_kind:x§]"
assert parse_prompt_mentions(prompt) == []
-def test_parse_skips_oversized_id_or_label():
+def test_parse_skips_oversized_id_or_label() -> None:
long_id = "x" * (MAX_MENTION_REF_ID_LENGTH + 1)
assert parse_prompt_mentions(f"[§skill:{long_id}§]") == []
@@ -66,10 +67,10 @@ def test_parse_skips_oversized_id_or_label():
# ── expand + scrub ────────────────────────────────────────────────────────────
-def test_expand_uses_resolver_and_degrades_unresolved_to_label_then_id():
+def test_expand_uses_resolver_and_degrades_unresolved_to_label_then_id() -> None:
prompt = "A [§skill:s1:Skill One§] B [§human:h1:EMAIL · DAVE§] C [§knowledge:k1§]"
- def resolver(mention):
+ def resolver(mention: PromptMention) -> str | None:
return "resolved-skill" if mention.kind == MentionKind.SKILL else None
expanded = expand_prompt_mentions(prompt, resolver)
@@ -77,29 +78,29 @@ def test_expand_uses_resolver_and_degrades_unresolved_to_label_then_id():
assert "[§" not in expanded
-def test_expand_scrubs_unknown_kind_tokens_but_keeps_legacy_forms():
+def test_expand_scrubs_unknown_kind_tokens_but_keeps_legacy_forms() -> None:
prompt = "x [§wat:id-1:Label§] y {{#context#}} z {{#node.var#}}"
- expanded = expand_prompt_mentions(prompt, lambda m: None)
+ expanded = expand_prompt_mentions(prompt, lambda _mention: None)
# unknown mention-shaped token degraded to its label; legacy forms untouched
assert expanded == "x Label y {{#context#}} z {{#node.var#}}"
-def test_scrub_degrades_colon_tokens_without_label_to_id_part():
+def test_scrub_degrades_colon_tokens_without_label_to_id_part() -> None:
assert scrub_mention_markers("see [§weird_kind:some-id§]") == "see some-id"
-def test_expand_empty_prompt_is_noop():
- assert expand_prompt_mentions("", lambda m: "x") == ""
+def test_expand_empty_prompt_is_noop() -> None:
+ assert expand_prompt_mentions("", lambda _mention: "x") == ""
-def test_extract_workflow_variable_selectors_supports_frontend_agent_task_format():
+def test_extract_workflow_variable_selectors_supports_frontend_agent_task_format() -> None:
selectors = extract_workflow_variable_selectors(
"Read {{#node-1.output#}}, compare {{#start.question#}}, leave {{#context#}} alone."
)
assert selectors == [("node-1", "output"), ("start", "question")]
-def test_extract_workflow_node_output_selectors_supports_current_frontend_markers_only():
+def test_extract_workflow_node_output_selectors_supports_current_frontend_markers_only() -> None:
selectors = extract_workflow_node_output_selectors(
"Read {{#node-1.output#}}, {{#sys.query#}}, [§node_output:legacy-node.report:PREV/report§], "
"and {{#node-1.output#}} again."
@@ -107,7 +108,7 @@ def test_extract_workflow_node_output_selectors_supports_current_frontend_marker
assert selectors == [("node-1", "output")]
-def test_workflow_previous_node_output_refs_from_selectors_materializes_refs():
+def test_workflow_previous_node_output_refs_from_selectors_materializes_refs() -> None:
refs = workflow_previous_node_output_refs_from_selectors([("node-1", "output"), ("node-2", "report", "url")])
assert refs == [
@@ -155,7 +156,7 @@ def soul() -> AgentSoulConfig:
)
-def test_soul_resolver_resolves_each_kind(soul: AgentSoulConfig):
+def test_soul_resolver_resolves_each_kind(soul: AgentSoulConfig) -> None:
resolver = build_soul_mention_resolver(soul)
prompt = (
"Use [§tool:tavily/tavily_search:tavily§], run [§cli_tool:ct-1:ffmpeg§], "
@@ -167,12 +168,12 @@ def test_soul_resolver_resolves_each_kind(soul: AgentSoulConfig):
assert expanded == ("Use tavily_search, run ffmpeg, ground in 产品手册, ask EMAIL · David Hayes.")
-def test_soul_resolver_unknown_ids_degrade(soul: AgentSoulConfig):
+def test_soul_resolver_unknown_ids_degrade(soul: AgentSoulConfig) -> None:
expanded = expand_prompt_mentions("[§knowledge:missing:旧产品手册§]", build_soul_mention_resolver(soul))
assert expanded == "旧产品手册"
-def test_soul_resolver_cli_tool_resolves_by_id_and_keeps_name_alias(soul: AgentSoulConfig):
+def test_soul_resolver_cli_tool_resolves_by_id_and_keeps_name_alias(soul: AgentSoulConfig) -> None:
resolver = build_soul_mention_resolver(soul)
# id is the contract; the name alias keeps tokens minted before ids existed working
assert expand_prompt_mentions("[§cli_tool:ct-1§]", resolver) == "ffmpeg"
@@ -193,7 +194,7 @@ def soul_with_provider_entry(soul: AgentSoulConfig) -> AgentSoulConfig:
return soul
-def test_soul_resolver_provider_all_tools_mention(soul_with_provider_entry: AgentSoulConfig):
+def test_soul_resolver_provider_all_tools_mention(soul_with_provider_entry: AgentSoulConfig) -> None:
resolver = build_soul_mention_resolver(soul_with_provider_entry)
# [§tool:<provider>/*§] = all tools of that provider
assert expand_prompt_mentions("Use [§tool:duckduckgo/*:DuckDuckGo 全部§].", resolver) == (
@@ -206,7 +207,7 @@ def test_soul_resolver_provider_all_tools_mention(soul_with_provider_entry: Agen
assert expand_prompt_mentions("[§tool:duckduckgo/*:DuckDuckGo 全部§]", bare) == "DuckDuckGo 全部"
-def test_soul_resolver_single_tool_resolves_via_provider_level_entry(soul_with_provider_entry: AgentSoulConfig):
+def test_soul_resolver_single_tool_resolves_via_provider_level_entry(soul_with_provider_entry: AgentSoulConfig) -> None:
# one tool offered through the provider-level ("all") entry still resolves
resolver = build_soul_mention_resolver(soul_with_provider_entry)
assert expand_prompt_mentions("[§tool:duckduckgo/ddg_search§]", resolver) == "ddg_search"
@@ -228,7 +229,7 @@ def node_job() -> WorkflowNodeJobConfig:
)
-def test_node_job_resolver_resolves_each_kind(node_job: WorkflowNodeJobConfig):
+def test_node_job_resolver_resolves_each_kind(node_job: WorkflowNodeJobConfig) -> None:
resolver = build_node_job_mention_resolver(node_job)
prompt = "Read [§node_output:start-1.tenders§] and produce [§output:qna_report§]; if unsure contact [§human:c-1§]."
@@ -244,7 +245,7 @@ def test_node_job_resolver_resolves_each_kind(node_job: WorkflowNodeJobConfig):
)
-def test_node_job_resolver_accepts_legacy_reversed_output_token(node_job: WorkflowNodeJobConfig):
+def test_node_job_resolver_accepts_legacy_reversed_output_token(node_job: WorkflowNodeJobConfig) -> None:
resolver = build_node_job_mention_resolver(node_job)
expanded = expand_prompt_mentions("[§qna_report:qna_report:output§]", resolver)
@@ -252,7 +253,7 @@ def test_node_job_resolver_accepts_legacy_reversed_output_token(node_job: Workfl
assert "final_output.qna_report" in expanded
-def test_node_job_resolver_exposes_system_text_output():
+def test_node_job_resolver_exposes_system_text_output() -> None:
node_job = WorkflowNodeJobConfig()
expanded = expand_prompt_mentions("[§output:text§]", build_node_job_mention_resolver(node_job))
@@ -260,7 +261,7 @@ def test_node_job_resolver_exposes_system_text_output():
assert expanded == "text (string)"
-def test_node_job_resolver_matches_ref_by_node_id_and_output_fields():
+def test_node_job_resolver_matches_ref_by_node_id_and_output_fields() -> None:
node_job = WorkflowNodeJobConfig.model_validate(
{"previous_node_output_refs": [{"node_id": "n-2", "output": "text"}]}
)
@@ -269,7 +270,7 @@ def test_node_job_resolver_matches_ref_by_node_id_and_output_fields():
assert expanded == "LLM/text"
-def test_normalize_previous_node_output_selector_returns_canonical_selector():
+def test_normalize_previous_node_output_selector_returns_canonical_selector() -> None:
assert normalize_previous_node_output_selector(
WorkflowPreviousNodeOutputRef(selector=["node-1", "report", "url"])
) == ("node-1", "report", "url")
@@ -283,7 +284,7 @@ def test_normalize_previous_node_output_selector_returns_canonical_selector():
# ── allowlists ────────────────────────────────────────────────────────────────
-def test_per_surface_allowlists_match_design():
+def test_per_surface_allowlists_match_design() -> None:
assert {
MentionKind.SKILL,
MentionKind.FILE,
diff --git a/api/tests/unit_tests/services/agent/test_skill_package_service.py b/api/tests/unit_tests/services/agent/test_skill_package_service.py
index 0a3a9ed696b..2d724ab170e 100644
--- a/api/tests/unit_tests/services/agent/test_skill_package_service.py
+++ b/api/tests/unit_tests/services/agent/test_skill_package_service.py
@@ -103,7 +103,7 @@ def test_normalized_highly_compressible_skill_passes_reinspection() -> None:
assert archive.read("data.bin") == payload
-def test_inspect_reports_uncompressed_size_without_rebuilding(monkeypatch: pytest.MonkeyPatch):
+def test_inspect_reports_uncompressed_size_without_rebuilding(monkeypatch: pytest.MonkeyPatch) -> None:
members = {"SKILL.md": _SKILL_MD.encode(), "scripts/run.py": b"print('hi')\n"}
def fail_if_called(*_args: object, **_kwargs: object) -> bytes:
@@ -141,7 +141,7 @@ def test_inspection_and_normalization_keep_distinct_size_meanings(prefix: str, c
assert inspection.uncompressed_size > normalized.manifest.size
-def test_shallowest_skill_md_preferred_during_normalization():
+def test_shallowest_skill_md_preferred_during_normalization() -> None:
manifest = _normalize({"SKILL.md": _SKILL_MD.encode(), "nested/SKILL.md": _SKILL_MD.encode()}).manifest
assert manifest.entry_path == "SKILL.md"
assert manifest.files == ["SKILL.md", "nested/SKILL.md"]
@@ -235,7 +235,7 @@ def test_normalization_preserves_selected_skill_contents(prefix: str, extra_memb
),
],
)
-def test_invalid_packages_rejected(members: dict[str, bytes], filename: str, code: str):
+def test_invalid_packages_rejected(members: dict[str, bytes], filename: str, code: str) -> None:
with pytest.raises(SkillPackageError) as exc_info:
_normalize(members, filename=filename)
assert exc_info.value.code == code
@@ -263,19 +263,19 @@ def test_opening_delimiter_with_trailing_text_is_not_frontmatter() -> None:
assert exc_info.value.code == "missing_skill_name"
-def test_non_zip_content_rejected():
+def test_non_zip_content_rejected() -> None:
with pytest.raises(SkillPackageError) as exc_info:
SkillPackageService().validate_and_normalize(content=b"not a zip", filename="skill.zip")
assert exc_info.value.code == "invalid_archive"
-def test_empty_archive_rejected():
+def test_empty_archive_rejected() -> None:
with pytest.raises(SkillPackageError) as exc_info:
SkillPackageService().validate_and_normalize(content=b"", filename="skill.zip")
assert exc_info.value.code == "empty_archive"
-def test_validate_and_normalize_rejects_skill_md_too_large(monkeypatch: pytest.MonkeyPatch):
+def test_validate_and_normalize_rejects_skill_md_too_large(monkeypatch: pytest.MonkeyPatch) -> None:
apply_config_overrides(monkeypatch, SKILL_PACKAGE_MAX_SKILL_MD_BYTES=8)
with pytest.raises(SkillPackageError) as exc_info:
@@ -283,7 +283,7 @@ def test_validate_and_normalize_rejects_skill_md_too_large(monkeypatch: pytest.M
assert exc_info.value.code == "skill_md_too_large"
-def test_validate_and_normalize_rejects_too_many_entries(monkeypatch: pytest.MonkeyPatch):
+def test_validate_and_normalize_rejects_too_many_entries(monkeypatch: pytest.MonkeyPatch) -> None:
apply_config_overrides(monkeypatch, SKILL_PACKAGE_MAX_ENTRIES=1)
with pytest.raises(SkillPackageError) as exc_info:
@@ -291,7 +291,7 @@ def test_validate_and_normalize_rejects_too_many_entries(monkeypatch: pytest.Mon
assert exc_info.value.code == "too_many_entries"
-def test_validate_and_normalize_rejects_archive_too_large_uncompressed(monkeypatch: pytest.MonkeyPatch):
+def test_validate_and_normalize_rejects_archive_too_large_uncompressed(monkeypatch: pytest.MonkeyPatch) -> None:
apply_config_overrides(monkeypatch, SKILL_PACKAGE_MAX_UNCOMPRESSED_BYTES=32)
with pytest.raises(SkillPackageError) as exc_info:
@@ -299,7 +299,7 @@ def test_validate_and_normalize_rejects_archive_too_large_uncompressed(monkeypat
assert exc_info.value.code == "archive_too_large"
-def test_validate_and_normalize_rejects_archive_too_large_uploaded_bytes(monkeypatch: pytest.MonkeyPatch):
+def test_validate_and_normalize_rejects_archive_too_large_uploaded_bytes(monkeypatch: pytest.MonkeyPatch) -> None:
apply_config_overrides(monkeypatch, UPLOAD_SKILL_FILE_SIZE_LIMIT=1)
with pytest.raises(SkillPackageError) as exc_info:
@@ -317,14 +317,16 @@ def test_high_compression_ratio_uses_absolute_expansion_limit(monkeypatch: pytes
assert exc_info.value.code == "archive_too_large"
-def test_validate_and_normalize_maps_member_decompression_failures_to_invalid_archive(monkeypatch: pytest.MonkeyPatch):
+def test_validate_and_normalize_maps_member_decompression_failures_to_invalid_archive(
+ monkeypatch: pytest.MonkeyPatch,
+) -> None:
original_read = zipfile.ZipFile.read
- def corrupted_read(self: zipfile.ZipFile, member: str | zipfile.ZipInfo, *args: object, **kwargs: object) -> bytes:
+ def corrupted_read(self: zipfile.ZipFile, member: str | zipfile.ZipInfo, pwd: bytes | None = None) -> bytes:
filename = member.filename if isinstance(member, zipfile.ZipInfo) else member
if filename == "scripts/run.py":
raise zlib.error("invalid distance too far back")
- return original_read(self, member, *args, **kwargs)
+ return original_read(self, member, pwd=pwd)
monkeypatch.setattr(zipfile.ZipFile, "read", corrupted_read)