openrouter.md

OpenRouter — .klickd v3.4 Integration

Docs: openrouter.ai/docs

OpenRouter provides a single OpenAI-compatible endpoint routing to 200+ models. Ideal for Soul Handoff workflows where the user switches AI providers mid-conversation — the .klickd file carries full context, and OpenRouter selects the target model.

import json
import httpx

with open("profile.klickd") as f:
    klickd = json.load(f)

system_prompt = klickd.get("user_preferences", "")
ctx = klickd.get("context", {})
if ctx.get("current_state"):
    system_prompt += f"\n\nCurrent state: {ctx['current_state']}"
if ctx.get("resume_trigger"):
    system_prompt += f"\n{ctx['resume_trigger']}"

# Strip benchmark/debug namespaces before injection
payload_keys_to_strip = [k for k in klickd if k.startswith("_")]
for k in payload_keys_to_strip:
    klickd.pop(k)

response = httpx.post(
    "https://openrouter.ai/api/v1/chat/completions",
    headers={
        "Authorization": "Bearer YOUR_OPENROUTER_API_KEY",
        "HTTP-Referer": "https://yourapp.com",
        "X-Title": "Your App Name"
    },
    json={
        "model": "qwen/qwen3-32b",   # or any OpenRouter model slug
        "max_tokens": 2048,
        "messages": [
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": "Let's continue where we left off."}
        ]
    }
)

print(response.json()["choices"][0]["message"]["content"])

Soul Handoff — cross-provider pattern:

1. User exports .klickd from Provider A (e.g. GPT-4o).
2. .klickd file is loaded by the new client.
3. user_preferences is injected as system message into Provider B via OpenRouter.
4. Companion identity (companion_identity.name, teaching_mode) is preserved.
5. Session resumes from context.resume_trigger.

Model selection: For injection_resistance_level: strict, prefer qwen/qwen3-32b or meta-llama/llama-3.3-70b-instruct.
Strip all _-prefixed fields before injection.