What the data for this category looks like
Data primarily comes from official product manuals, marketing promotional materials, specification documents, sales training materials, and offline exhibition organized materials of construction machinery manufacturers. Updates are triggered by new product launches, marketing campaign launches, or parameter adjustments, with no fixed schedule. Most documents combine structured parameter tables and long-form explanatory text, with fields including equipment model, rated power, working radius, total machine weight, applicable working conditions, and more. Most units follow international standard units, with some scenarios using non-standard custom units.
What constraints these characteristics impose on vector models and indexing
Documents contain a large volume of structured parameters and long-form explanatory text. This requires vector models to support both structured field encoding and long-context encoding. This prevents loss of parameter information or truncation of long text. Updates follow no fixed schedule, and single-update data volume fluctuates significantly. This requires indexes to support incremental updates and dynamic scaling. This avoids performance loss caused by full index reconstruction. Mixed standardized and non-standard units exist across fields. Unit information must be retained during indexing for vector encoding. This ensures accurate semantic matching during retrieval. Marketing materials include a large number of similar phrasing. This requires indexes to support merging similar content. This reduces retrieval redundancy and improves retrieval efficiency.
How to set the configurations
| Configuration Item | Recommended Approach | Basis for This Selection |
|---|---|---|
embedding_model | text-embedding-v3 | Supports long-context encoding and structured field semantic extraction, adapts to the mixed content structure of construction machinery documents |
chunk_size | 800–1200 characters | Balances encoding integrity for parameter tables and long-form explanatory text in construction machinery documents, avoids semantic loss caused by overly long single segments |
recall_top_k | Top 8–12 results | Covers multi-dimensional equipment parameters and scenario requirements, while controlling computational load during the retrieval phase |
rerank_top_k | Top 3–5 results | Filters redundant recall results, improves response speed for marketing content retrieval, and matches precise matching requirements |
embedding_api_base | Fill in according to actual gateway configuration | Adapts to third-party API call links, resolves cross-region or channel restriction issues |
enable_incremental_index | Enabled | Adapts to the non-fixed update cycle characteristic of construction machinery data, reduces performance loss from full index reconstruction |
The parameter values provided on this page are all common starting points for configuration setup. Actual values are affected by material form, data volume, and business rules. Specific issues require targeted analysis. It is recommended to test on your own samples before finalizing settings.
Three common mistakes
- Phenomenon: A "no available channel" prompt appears when calling the embedding model, and the interface returns a 403 status code. Cause: The
embedding_api_baseparameter is not configured correctly, or the third-party API gateway has not activated the call permission for the corresponding model. - Phenomenon: Timeouts occur during the re-ranking retrieval link of knowledge base question answering, and the response duration exceeds expectations. Cause: The
chunk_sizevalue is too small, leading to an excessive number of segments, or therecall_top_ksetting is too high, increasing the computational load of vector retrieval and re-ranking. - Phenomenon: Equipment parameters in retrieval results do not match marketing scenarios, and core field information is missing. Cause: Unit information is not retained for vector encoding, or the selected embedding model cannot effectively adapt to the mixed content of structured parameters and long text.
How to confirm the configuration is correct
- Access the FastGPT embedding model configuration interface, confirm that the
embedding_modelandembedding_api_baseparameters are filled in correctly, initiate a single test call, and verify that vector data can be generated normally without errors. - Upload a construction machinery product parameter document, check whether the automatically segmented content fully retains equipment models, parameter values and units, with no forced truncation or unreasonable splitting.
- Initiate a retrieval test based on marketing keywords, verify that the number of recall results and the number of outputs after re-ranking conform to the preset configuration rules.
- Upload an updated marketing material, confirm that the index automatically triggers incremental updates, with no need for manual full index reconstruction operations.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.