Video intelligence with Q-learning — analyze sales calls with 3-channel Gemini multimodal analysis, store insights in Qdrant with Q-learning feedback
npx @senso-ai/shipables install anthroos/callmindUse this skill to analyze sales call recordings and retrieve Q-learning-ranked insights via the CallMind API.
CallMind must be running (locally or remote). Set the base URL:
CALLMIND_URL=http://localhost:8000
Get an API key by registering:
curl -X POST $CALLMIND_URL/register -d "username=your_name"
Upload a video file or YouTube URL for 3-channel multimodal analysis:
# File upload
curl -X POST $CALLMIND_URL/api/upload \
-H "Authorization: Bearer YOUR_API_KEY" \
-F "client_name=Acme Corp" \
-F "video_file=@recording.mp4"
# YouTube URL
curl -X POST $CALLMIND_URL/api/upload \
-H "Authorization: Bearer YOUR_API_KEY" \
-F "client_name=Acme Corp" \
-F "youtube_url=https://youtube.com/watch?v=..."
Response includes a job_id. Poll status:
curl $CALLMIND_URL/api/status/{job_id}
Retrieve insights ranked by Q-value (hybrid: 60% semantic + 40% Q-value):
curl "$CALLMIND_URL/api/client/{client_id}/insights" \
-H "Authorization: Bearer YOUR_API_KEY"
Each insight includes:
type: pain_point, objection, need, decision_maker, budget, timeline, competitor, next_step, sentiment, relationshipchannel: text (transcript), visual (body language), fusion (cross-modal)q_value: learned importance score (-0.5 to 1.0)content: the insight textaction_point: concrete recommended actionInsights learn from deal outcomes. After a deal closes:
High Q-value insights float to the top of pre-call briefings.