User Profiles

import requests

# GET /v1/profile
# Flat convenience fields (name/timezone/language) live at the top
# level; all other facts sit under the "profile.static" and
# "profile.dynamic" buckets.
r = requests.get(
    "https://api.hebbrix.com/v1/profile",
    headers={"Authorization": "Bearer <your-api-key>"},
)
data = r.json()

print(f"Name: {data.get('name')}")
print(f"Timezone: {data.get('timezone')}")
print(f"Language: {data.get('language')}")

# Every other fact lives under profile.static (long-lived) or
# profile.dynamic (short-lived). Each entry is {key, value, confidence, ...}.
for fact in data["profile"]["static"]:
    print(f"  static · {fact['key']} = {fact['value']}")
for fact in data["profile"]["dynamic"]:
    print(f"  dynamic · {fact['key']} = {fact['value']}")

import requests

# POST /v1/profile/dialectic — ask natural-language questions about the user.
# Send the prompt as `user_message` (preferred); legacy `message` / `question`
# are still accepted.
r = requests.post(
    "https://api.hebbrix.com/v1/profile/dialectic",
    headers={"Authorization": "Bearer <your-api-key>"},
    json={"user_message": "What are this user's main professional interests?"},
)
body = r.json()

# Preferred response fields
print(body["assistant_message"])   # the LLM answer
print(body["session_id"])          # echo back in the next turn for continuity

# Legacy alias (same value as assistant_message)
# body["response"] is also available for older integrations.

# Continue the conversation — reuse session_id to link turns
r2 = requests.post(
    "https://api.hebbrix.com/v1/profile/dialectic",
    headers={"Authorization": "Bearer <your-api-key>"},
    json={
        "user_message": "Tell me more about how they like to communicate.",
        "session_id": body["session_id"],
    },
)

import requests

# PATCH /v1/profile — bulk-update profile facts. Body is a flat
# {fact_key: fact_value} dict. Existing facts are updated; new ones
# are created as "static" / "other" automatically.
r = requests.patch(
    "https://api.hebbrix.com/v1/profile",
    headers={"Authorization": "Bearer <your-api-key>"},
    json={
        "name": "John Doe",
        "timezone": "UTC",
        "language": "en",
        "role": "Senior Engineer",
        "company": "Acme Inc",
    },
)

body = r.json()
print(body["status"])     # "success"
print(body["updated"])    # count of facts updated in place
print(body["created"])    # count of new facts created
print(body["message"])    # e.g. "Updated 3, created 2 profile facts"

import os
import requests

BASE = "https://api.hebbrix.com/v1"
H = {"Authorization": f"Bearer {os.environ['HEBBRIX_API_KEY']}"}

# POST /v1/chat/completions — set features.user_profile=True to have
# the backend inject profile facts into the context automatically.
r = requests.post(
    f"{BASE}/chat/completions",
    headers=H,
    json={
        "model": "gpt-5-nano",
        "messages": [
            {"role": "user", "content": "Recommend a book for me"},
        ],
        "features": {
            "memory": True,
            "user_profile": True,  # inject profile context
        },
    },
)

# The assistant sees the caller's profile and gives a personalized answer
print(r.json()["choices"][0]["message"]["content"])

import requests

BASE = "https://api.hebbrix.com/v1"
H = {"Authorization": "Bearer <your-api-key>"}

# Each end-user in your SaaS gets their own collection id.
COLLECTION = "coll_123"

# 1) PREVIEW — dry_run returns the planned changes WITHOUT persisting.
preview = requests.patch(
    f"{BASE}/profile",
    headers=H,
    json={
        "role": "Product Manager",
        "timezone": "America/New_York",
        "collection_id": COLLECTION,  # reserved key — scopes the facts
        "dry_run": True,              # reserved key — preview only
    },
).json()

print(preview["status"])         # "preview"
print(preview["would_update"])   # N facts that would be updated in place
print(preview["would_create"])   # N facts that would be created
print(preview["collection_id"])  # "coll_123"
print(preview["changes"])        # [...] per-fact diff, nothing written yet

# 2) REAL WRITE — drop dry_run (or set it False) to persist the facts,
#    tagged to this end-user's collection so they stay isolated.
written = requests.patch(
    f"{BASE}/profile",
    headers=H,
    json={
        "role": "Product Manager",
        "timezone": "America/New_York",
        "collection_id": COLLECTION,
    },
).json()
print(written["status"])   # "success"
print(written["updated"], written["created"])

# 3) SCOPED READ — only facts for this collection come back; another
#    end-user's profile never leaks in.
profile = requests.get(
    f"{BASE}/profile",
    headers=H,
    params={"collection_id": COLLECTION},
).json()

for fact in profile["profile"]["static"]:
    print(f"  {fact['key']} = {fact['value']}")