Application Context

Endpoints for managing application context — the domain briefing that guides the LLM compiler when resolving user intent into concept and condition definitions. Context is versioned and takes effect immediately on creation. No server restart required.

---

Create Context

POST /context

Creates a new application context version and immediately activates it. Deactivates the previous version atomically. The LLM compiler uses the active context for all subsequent task compilations.

Request — CreateContextRequest

Field	Type	Required	Description
domain	dict	Required	Domain description object. Must include description (str). Optional: entities (list), decisions (list[str]).
domain.description	str	Required	Natural language description of what the application does and what domain it operates in.
domain.entities	list[dict]	Optional	List of entity declarations. Each has name (str) and description (str).
domain.decisions	list[str]	Optional	Decision type names relevant to this domain (e.g. churn_risk, fraud_alert).
behavioural	dict	Optional	Behavioural settings. See fields below.
behavioural.data_cadence	str	Optional	'batch' | 'streaming' | 'mixed'. Default 'batch'.
behavioural.meaningful_windows	dict	Optional	{ 'min': str, 'max': str } — e.g. { 'min': '30d', 'max': '90d' }. Informs calibration window clamping.
behavioural.regulatory	list[str]	Optional	Regulatory frameworks in scope (e.g. ['GDPR', 'SOC2', 'HIPAA']).
semantic_hints	list[dict]	Optional	Domain-specific term definitions. Each has term (str) and definition (str).
calibration_bias	dict	Optional	Calibration cost asymmetry. See fields below.
calibration_bias.false_negative_cost	str	Optional	'high' | 'medium' | 'low'. Cost of missing a true positive.
calibration_bias.false_positive_cost	str	Optional	'high' | 'medium' | 'low'. Cost of a false alarm.

bias_direction is auto-derived and cannot be set manually. false_negative_cost > false_positive_cost → recall. Reverse → precision. Equal → balanced.

Response — ApplicationContext

Field	Type	Description
context_id	str (UUID)	Auto-generated unique identifier.
version	str	Auto-assigned version string: v1, v2, v3...
domain	dict	The domain context as submitted.
behavioural	dict | None	The behavioural context as submitted.
semantic_hints	list	The semantic hints as submitted.
calibration_bias	dict | None	The calibration bias including auto-derived bias_direction.
created_at	str	ISO 8601 UTC timestamp.
is_active	bool	True — newly created versions are always immediately active.

Response Codes

Status	Description
201	Context created and active.
400	Validation error — missing required field or invalid enum value.
401	Unauthorised — missing or invalid X-API-Key.

Python Example

import memintel

client = memintel.AsyncClient()

context = await client.context.create(
    domain={
        "description": "B2B SaaS churn detection for mid-market software companies.",
        "entities": [
            {"name": "account", "description": "company-level subscription"},
            {"name": "user",    "description": "individual platform user"},
        ],
        "decisions": ["churn_risk", "expansion_opportunity"],
    },
    behavioural={
        "data_cadence": "batch",
        "meaningful_windows": {"min": "30d", "max": "90d"},
        "regulatory": ["GDPR", "SOC2"],
    },
    semantic_hints=[
        {"term": "active user",        "definition": "logged in AND performed core action in last 14 days"},
        {"term": "high value account", "definition": "ARR above $50,000"},
    ],
    calibration_bias={
        "false_negative_cost": "high",
        "false_positive_cost": "medium",
    },
)

print(context.version)    # v1
print(context.is_active)  # True
# context.calibration_bias.bias_direction → 'recall' (auto-derived)

---

Get Active Context

GET /context

Returns the currently active context version. Returns HTTP 404 if no context has been defined.

Response Codes

Status	Description
200	Active ApplicationContext returned.
401	Unauthorised.
404	No active context defined. error.type: not_found.

Python Example

try:
    context = await client.context.get_active()
    print(context.version)      # v2
    print(context.is_active)    # True
except memintel.NotFoundError:
    print("No context defined yet")

---

List Context Versions

GET /context/versions

Returns all context versions, newest first. Useful for auditing the history of domain definition changes.

Response

list[ApplicationContext] ordered by created_at descending.

Python Example

versions = await client.context.list_versions()

for v in versions:
    print(f"{v.version} — {'active' if v.is_active else 'inactive'} — {v.created_at}")
# v2 — active   — 2026-03-27T10:00:00Z
# v1 — inactive — 2026-03-01T09:00:00Z

---

Get Context Version

GET /context/versions/{version}

Returns a specific context version by version string.

Path Parameters

Field	Type	Required	Description
version	str	Required	Version string — e.g. 'v1', 'v2'.

Response Codes

Status	Description
200	ApplicationContext for the requested version.
401	Unauthorised.
404	Version not found.

Python Example

v1 = await client.context.get_version("v1")
print(v1.domain["description"])
print(v1.is_active)  # False — superseded by v2

---

Get Context Impact

GET /context/impact

Returns a summary of how many tasks were compiled under older context versions. Use after updating context to identify tasks that would benefit from recompilation.

Response — ContextImpactResult

Field	Type	Description
current_version	str	The currently active context version.
tasks_on_current_version	int	Tasks compiled under the current version.
tasks_on_older_versions	list[dict]	List of { version, task_count } for each older version with active tasks.
total_stale_tasks	int	Total tasks not on the current context version.

Response Codes

Status	Description
200	ContextImpactResult.
401	Unauthorised.

Python Example

impact = await client.context.get_impact()

print(f"Current version: {impact.current_version}")
print(f"Tasks on current: {impact.tasks_on_current_version}")
print(f"Stale tasks: {impact.total_stale_tasks}")

for entry in impact.tasks_on_older_versions:
    print(f"  {entry['version']}: {entry['task_count']} tasks need recompilation")