Configuration Presets

KAPEX allows you to tune memory behavior on a per-tenant basis using named presets. Each preset bundles parameters for decay, retrieval, token budgets, and signal weighting -- optimized for a specific domain.

GET /profile

Retrieve the active configuration for your tenant.

curl https://api.getkapex.ai/api/v1/profile \
  -H "X-API-Key: your_api_key_here"

PUT /profile

Apply a preset or set custom overrides.

Apply a Preset

curl -X PUT https://api.getkapex.ai/api/v1/profile \
  -H "X-API-Key: your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "preset": "reflective"
  }'

Custom Overrides

curl -X PUT https://api.getkapex.ai/api/v1/profile \
  -H "X-API-Key: your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "updates": {
      "decay_profile": {
        "base_rate": 0.01,
        "processing_boost": 0.5
      },
      "injection_template": {
        "max_tokens": 8000
      }
    }
  }'

List Available Presets

curl https://api.getkapex.ai/api/v1/profile/presets \
  -H "X-API-Key: your_api_key_here"

Available Presets

KAPEX provides 16 domain-specific presets. Each tunes decay rates, token budgets, signal weights, and retrieval behavior for its target use case.

Core Presets

`reflective`

Optimized for journaling, therapy companions, and personal reflection apps. Slow decay preserves important disclosures. Larger token budget gives the LLM more context to work with.

Best for: Mental health platforms, therapy companions, journaling apps, personal reflection tools.

`customer_service`

High recency bias with faster decay. Recent interactions stay salient while older conversations fade quickly. Focused token budget keeps responses concise and case-relevant.

Best for: Help desks, customer support bots, ticket-based interactions.

`education`

Tuned for learning platforms. Active study sessions (processing events) create clear differentiation between practiced and unpracticed material.

Best for: Tutoring systems, learning platforms, spaced-repetition applications.

`healthcare`

Conservative configuration with higher residual floors. Medical information persists longer and decays more slowly. Safety checks run at maximum intensity.

Best for: Clinical documentation, patient context, care coordination, medical assistants.

`sales`

Assertive framing with emphasis on factual recall. Optimized for remembering deal context, customer preferences, and relationship history.

Best for: CRM assistants, sales engagement, account management tools.

`legal`

Precision-oriented with minimal emotional weighting. Facts and dates persist; subjective impressions decay faster. Low injection threshold ensures comprehensive context retrieval.

Best for: Legal research assistants, case management, compliance tools.

All Presets

Custom Configuration

You can override individual parameters without applying a preset. The updates object supports nested configuration sections:

Decay Profile

Controls how quickly memories lose salience over time.

{
  "updates": {
    "decay_profile": {
      "base_rate": 0.02,
      "processing_boost": 0.3,
      "residual_floor": 0.05
    }
  }
}

Injection Template

Controls how memories are assembled into the LLM context block.

{
  "updates": {
    "injection_template": {
      "max_tokens": 6000,
      "injection_threshold": 0.25
    }
  }
}

Governance Floors

KAPEX enforces minimum values on safety-critical parameters. You cannot configure the system to forget things too aggressively or suppress memories below safe thresholds.

If a requested value falls below the governance floor, the system clamps it to the floor value.

Python Example

import requests

API_KEY = "your_api_key_here"
BASE = "https://api.getkapex.ai/api/v1"
HEADERS = {"X-API-Key": API_KEY, "Content-Type": "application/json"}

# Apply a domain preset
resp = requests.put(
    f"{BASE}/profile",
    headers=HEADERS,
    json={"preset": "healthcare"}
)
print(f"Profile updated: {resp.json()}")

# Or use custom overrides
resp = requests.put(
    f"{BASE}/profile",
    headers=HEADERS,
    json={
        "updates": {
            "decay_profile": {"base_rate": 0.008, "processing_boost": 0.6},
            "injection_template": {"max_tokens": 8000, "injection_threshold": 0.15}
        }
    }
)
print(f"Custom config applied: {resp.json()}")