What the Data for This Use Case Looks Like
Data for meeting minutes organization comes from three primary sources: meeting audio transcription files, written speech submissions from attendees, and original check-in and speech records exported from meeting management systems. Updates occur either as batch generation after a single meeting concludes, or real-time synchronization. Single document size varies based on meeting duration. Document structure includes four core field groups: core meeting details (topic, time, participating parties), segmented speech content, resolution item entries, and follow-up task lists. Most fields are text-based. Some are identification fields linked to attendee IDs, meeting numbers, and similar identifiers.
Constraints During Deployment and Upgrade
Long-form transcribed meeting speech content leads to large single-data volume. Adjust storage node cache thresholds during deployment to avoid memory overflow. Multi-source data access requires configuring multi-format file parsing adapters. Ensure compatibility with newly exported meeting system formats during upgrades. Mixed storage of structured fields and unstructured speech text requires configuring associated mappings between relational databases and vector databases, to prevent data association failures. Real-time synchronized meeting data requires deploying asynchronous task processing pipelines. Ensure task queue concurrency limits match the number of concurrent meetings during upgrades. Meeting data containing sensitive information requires configuring compliant identity authentication rules, to ensure deployment meets industry regulatory requirements.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
MONGODB_URI | mongodb://username:password@host:port/dbname?authSource=admin | Stores unstructured speech text and associated identification fields for meeting minutes, supports high-concurrency read/write operations |
PG_URL | postgresql://user:pass@host:5432/fastgpt?schema=public | Stores standardized fields for meeting minutes (such as meeting topic, resolution items), supports complex associated queries |
REDIS_URL | redis://:password@host:6379/0 | Caches temporary meeting transcription files and status identifiers for asynchronous task queues, reduces database read/write pressure |
DB_MAX_LINK | 200–300 | Adapts to database connection requirements for multiple concurrent meetings, prevents connection pool exhaustion |
STORAGE_S3_ENDPOINT | Fill in the intranet endpoint provided by the cloud service provider | Stores large-volume attachments such as original meeting audio and transcription files, reduces public network transmission costs |
FILE_TOKEN_KEY | Randomly generated 32-bit string | Encrypts access tokens for meeting files, prevents unauthorized access to sensitive meeting data |
Three Common Mistakes
- Symptom: Importing meeting audio transcription files after online deployment results in garbled text, while local deployment operates normally. Cause: Intranet access permissions for
STORAGE_S3_ENDPOINTwere not configured correctly, leading to encoding loss during file transfer. - Symptom: After upgrading to FastGPT v4.17.0, calling the meeting minutes organization node in a workflow fails to retrieve speech content output by code runs. Cause: Associated database table structures were not updated synchronously, leading to failed mapping between structured fields and unstructured text.
- Symptom: Unauthorized external requests can directly access sensitive meeting data when calling the meeting minutes organization function. Cause:
FILE_TOKEN_KEYorROOT_KEYwas not configured, leading to missing identity authentication checks.
How to Verify Proper Configuration
- The built-in database connectivity detection tool is run to verify the connection status of
MONGODB_URI,PG_URL, andREDIS_URL, and confirm that the connection pool concurrency limit meets expected requirements. - A single meeting transcription test file is uploaded, the file storage path is checked against the
STORAGE_S3_ENDPOINTconfiguration, and no transfer error logs are present. - A batch simulated meeting data import task is initiated, the execution progress of the asynchronous task queue is checked, and it is confirmed that the
DB_MAX_LINKconfiguration does not trigger connection limit exceeded errors. - An API request without a valid token is constructed, and it is confirmed that the
FILE_TOKEN_KEYandROOT_KEYconfiguration blocks unauthorized access.
The values provided are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-28.