feat: add governed knowledge lifecycle

This commit is contained in:
hailin
2026-07-27 19:01:19 +08:00
parent 1a49ce4eae
commit 07c9ea1cc4
32 changed files with 5893 additions and 178 deletions

View File

@@ -1,5 +1,8 @@
import uuid
from django.conf import settings
from django.db import models
from pgvector.django import VectorField
class KnowledgeBase(models.Model):
@@ -24,6 +27,7 @@ class KnowledgeBase(models.Model):
class KnowledgeDocument(models.Model):
public_id = models.UUIDField(default=uuid.uuid4, unique=True, editable=False)
knowledge_base = models.ForeignKey(
KnowledgeBase,
on_delete=models.CASCADE,
@@ -31,10 +35,19 @@ class KnowledgeDocument(models.Model):
)
file_name = models.CharField(max_length=255)
file_type = models.CharField(max_length=80, blank=True)
extension = models.CharField(max_length=16, blank=True)
mime_type = models.CharField(max_length=160, blank=True)
file_size = models.PositiveBigIntegerField(default=0)
sha256 = models.CharField(max_length=64, blank=True)
storage_path = models.CharField(max_length=500, blank=True)
content_text = models.TextField(blank=True)
scan_status = models.CharField(max_length=40, default="pending")
scan_result = models.JSONField(default=dict, blank=True)
parse_status = models.CharField(max_length=40, default="pending")
parse_error = models.TextField(blank=True)
processing_attempts = models.PositiveSmallIntegerField(default=0)
task_id = models.CharField(max_length=120, blank=True)
indexed_at = models.DateTimeField(null=True, blank=True)
created_by = models.ForeignKey(
settings.AUTH_USER_MODEL,
null=True,
@@ -46,6 +59,14 @@ class KnowledgeDocument(models.Model):
class Meta:
ordering = ["-created_at"]
indexes = [
models.Index(fields=["knowledge_base", "parse_status", "created_at"]),
models.Index(fields=["sha256"]),
]
@property
def has_original(self) -> bool:
return bool(self.storage_path)
class KnowledgeItem(models.Model):
@@ -57,6 +78,9 @@ class KnowledgeItem(models.Model):
title = models.CharField(max_length=200)
problem = models.TextField()
applicable_scope = models.TextField(blank=True)
category_code = models.CharField(max_length=80, blank=True)
related_system = models.CharField(max_length=120, blank=True)
environment_scope = models.CharField(max_length=120, blank=True)
error_code = models.CharField(max_length=120, blank=True)
root_cause = models.TextField(blank=True)
troubleshooting_steps = models.TextField(blank=True)
@@ -86,6 +110,221 @@ class KnowledgeItem(models.Model):
ordering = ["-updated_at"]
class TroubleshootingPlan(models.Model):
knowledge_base = models.ForeignKey(
KnowledgeBase,
on_delete=models.CASCADE,
related_name="troubleshooting_plans",
)
title = models.CharField(max_length=200)
problem_pattern = models.TextField()
applicable_scope = models.TextField(blank=True)
category_code = models.CharField(max_length=80, blank=True)
related_system = models.CharField(max_length=120, blank=True)
environment_scope = models.CharField(max_length=120, blank=True)
decision_rules = models.TextField(blank=True)
troubleshooting_steps = models.TextField()
risk_note = models.TextField(blank=True)
status = models.CharField(max_length=40, default="draft")
source_ticket = models.ForeignKey(
"tickets.Ticket",
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="generated_troubleshooting_plans",
)
created_by = models.ForeignKey(
settings.AUTH_USER_MODEL,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="created_troubleshooting_plans",
)
reviewed_by = models.ForeignKey(
settings.AUTH_USER_MODEL,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="reviewed_troubleshooting_plans",
)
created_at = models.DateTimeField(auto_now_add=True)
updated_at = models.DateTimeField(auto_now=True)
class Meta:
ordering = ["-updated_at"]
class GovernedAssetType(models.TextChoices):
KNOWLEDGE_ITEM = "knowledge_item", "知识"
TROUBLESHOOTING_PLAN = "troubleshooting_plan", "排查方案"
SOLUTION = "solution", "解决方案"
class GovernedAsset(models.Model):
public_id = models.UUIDField(default=uuid.uuid4, unique=True, editable=False)
asset_type = models.CharField(max_length=40, choices=GovernedAssetType.choices)
object_id = models.PositiveBigIntegerField()
current_published_version = models.ForeignKey(
"GovernedAssetVersion",
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="+",
)
created_by = models.ForeignKey(
settings.AUTH_USER_MODEL,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="created_governed_assets",
)
created_at = models.DateTimeField(auto_now_add=True)
updated_at = models.DateTimeField(auto_now=True)
class Meta:
constraints = [
models.UniqueConstraint(
fields=["asset_type", "object_id"],
name="unique_governed_asset_object",
)
]
indexes = [models.Index(fields=["asset_type", "object_id"])]
class GovernedAssetVersionStatus(models.TextChoices):
DRAFT = "draft", "草稿"
PENDING_REVIEW = "pending_review", "待审核"
REJECTED = "rejected", "已驳回"
PUBLISHED = "published", "已发布"
OFFLINE = "offline", "已下线"
class GovernedAssetVersion(models.Model):
asset = models.ForeignKey(
GovernedAsset,
on_delete=models.CASCADE,
related_name="versions",
)
version = models.PositiveIntegerField()
status = models.CharField(
max_length=40,
choices=GovernedAssetVersionStatus.choices,
default=GovernedAssetVersionStatus.DRAFT,
)
content = models.JSONField(default=dict)
change_summary = models.CharField(max_length=500, blank=True)
source_version = models.ForeignKey(
"self",
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="derived_versions",
)
created_by = models.ForeignKey(
settings.AUTH_USER_MODEL,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="created_asset_versions",
)
submitted_by = models.ForeignKey(
settings.AUTH_USER_MODEL,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="submitted_asset_versions",
)
reviewed_by = models.ForeignKey(
settings.AUTH_USER_MODEL,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="reviewed_asset_versions",
)
review_comment = models.TextField(blank=True)
created_at = models.DateTimeField(auto_now_add=True)
updated_at = models.DateTimeField(auto_now=True)
submitted_at = models.DateTimeField(null=True, blank=True)
reviewed_at = models.DateTimeField(null=True, blank=True)
published_at = models.DateTimeField(null=True, blank=True)
offline_at = models.DateTimeField(null=True, blank=True)
class Meta:
ordering = ["-version"]
constraints = [
models.UniqueConstraint(
fields=["asset", "version"],
name="unique_governed_asset_version",
),
models.UniqueConstraint(
fields=["asset"],
condition=models.Q(status=GovernedAssetVersionStatus.DRAFT),
name="unique_draft_governed_asset_version",
),
models.UniqueConstraint(
fields=["asset"],
condition=models.Q(status=GovernedAssetVersionStatus.PENDING_REVIEW),
name="unique_pending_governed_asset_version",
),
]
indexes = [
models.Index(fields=["status", "updated_at"]),
models.Index(fields=["asset", "status", "version"]),
]
class TicketKnowledgeDistillation(models.Model):
ticket = models.ForeignKey(
"tickets.Ticket",
on_delete=models.CASCADE,
related_name="knowledge_distillations",
)
run = models.OneToOneField(
"agents.AgentRun",
on_delete=models.PROTECT,
related_name="ticket_knowledge_distillation",
)
references = models.JSONField(default=list)
retrieval_policy = models.JSONField(default=dict)
should_distill = models.BooleanField(null=True, blank=True)
reason = models.TextField(blank=True)
assessment = models.JSONField(default=dict, blank=True)
knowledge_item = models.ForeignKey(
KnowledgeItem,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="distillation_records",
)
troubleshooting_plan = models.ForeignKey(
TroubleshootingPlan,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="distillation_records",
)
solution = models.ForeignKey(
"solutions.Solution",
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="distillation_records",
)
processed_at = models.DateTimeField(null=True, blank=True)
created_by = models.ForeignKey(
settings.AUTH_USER_MODEL,
null=True,
blank=True,
on_delete=models.SET_NULL,
related_name="created_knowledge_distillations",
)
created_at = models.DateTimeField(auto_now_add=True)
class Meta:
ordering = ["-created_at"]
indexes = [models.Index(fields=["ticket", "created_at"])]
class KnowledgeChunk(models.Model):
knowledge_base = models.ForeignKey(
KnowledgeBase,
@@ -106,15 +345,37 @@ class KnowledgeChunk(models.Model):
on_delete=models.CASCADE,
related_name="chunks",
)
troubleshooting_plan = models.ForeignKey(
TroubleshootingPlan,
null=True,
blank=True,
on_delete=models.CASCADE,
related_name="chunks",
)
historical_ticket = models.ForeignKey(
"tickets.Ticket",
null=True,
blank=True,
on_delete=models.CASCADE,
related_name="knowledge_chunks",
)
chunk_index = models.PositiveIntegerField(default=0)
content = models.TextField()
embedding = models.JSONField(default=list, blank=True)
embedding_vector = VectorField(dimensions=1536, null=True, blank=True)
embedding_model = models.CharField(max_length=120, default="local-hash-v1")
metadata = models.JSONField(default=dict, blank=True)
created_at = models.DateTimeField(auto_now_add=True)
class Meta:
ordering = ["knowledge_base_id", "document_id", "item_id", "chunk_index"]
ordering = [
"knowledge_base_id",
"document_id",
"item_id",
"troubleshooting_plan_id",
"historical_ticket_id",
"chunk_index",
]
indexes = [
models.Index(fields=["knowledge_base", "chunk_index"]),
]