What data looks like for this category
The knowledge base data for power generation equipment manufacturing mainly comes from internal enterprise R&D design documents, operation and maintenance manuals, parts specifications, and industry technical standard documents. The update rhythm is adjusted with the launch of new models, revision of industry standards, or accumulation of operation and maintenance experience. Documents include two types of content: structured parameter tables and unstructured operation instructions. Fields cover equipment model, rated capacity, operating temperature range, fault codes, etc. Units are mostly megawatts, revolutions per minute, degrees Celsius, etc. The length of individual documents ranges from hundreds of characters for fault code descriptions to tens of thousands of characters for complete machine operation and maintenance manuals.
What constraints these characteristics bring to the "knowledge base retrieval and recall" link
Structured parameters in documents require retrieval to support both keyword matching and numerical range matching, to avoid retrieving irrelevant content with mismatched parameter units or values. The length span of unstructured documents is large, so chunking and retrieval need to adapt to text of different lengths, preventing long texts from being overly truncated or short texts failing to cover complete semantics. Data updates have no fixed cycle, so a flexible incremental synchronization mechanism is needed to ensure that new model documents are included in the knowledge base in a timely manner. Some internal operation and maintenance documents involve business details, so permission control must be combined during retrieval to prevent sensitive content from being accessed without authorization.
How to set the configuration
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.72–0.85 | Power generation equipment parameter documents require high matching accuracy to avoid retrieving irrelevant content with mismatched parameter units or values |
limit | Top 8–12 entries | The number of relevant retrieval results for a single operation and maintenance document should not be too large, to avoid context overflow affecting subsequent generation effects |
searchMode | Hybrid retrieval | Balances keyword matching and precise matching of numerical parameters, adapting to both structured and unstructured content in documents |
usingReRank | Enabled | Performs secondary ranking on retrieval results for long-text operation and maintenance documents, improving the relevance of final retrieved content |
datasetSearchUsingExtensionQuery | Enabled | Expands user consultation questions to supplement power generation equipment-specific keywords such as "rated power" and "rotational speed", adapting to parameter retrieval needs |
collectionIds | Specify knowledge base collection IDs related to power generation equipment business | Only retrieve documents for the target business, excluding knowledge base content from unrelated businesses |
Three common mistakes
- Symptom: Chinese file names display garbled characters when uploading knowledge base files via API. Cause: The correct character encoding format was not specified in the API request header, leading to encoding errors during transmission of the file name parameter.
- Symptom: Unable to select a locally deployed ChatGLM2 model when creating a knowledge base, even after connecting via OneAPI. Cause: The custom model channel corresponding to OneAPI was not added in the FastGPT v4.17.0 model configuration page, or the configured model name did not match the model identifier returned by OneAPI.
- Symptom: Unable to properly crawl page content when importing a Notion link as a knowledge base source. Cause: The API access permission for Notion was not configured, or the link did not have public/team access permissions enabled, resulting in failure to obtain page data.
How to confirm the configuration is correct
- Upload a single power generation equipment operation and maintenance document as test data, initiate a consultation containing "fault code" and "maintenance steps", and check whether the retrieved results include the target content.
- View the retrieval statistics panel to confirm that the number of retrieval results matches the setting of the
limitparameter. - Test the API upload function by uploading a test file with a Chinese file name, and confirm that the file name is displayed without garbled characters.
- Check the associated collections of the knowledge base to confirm that only
collectionIdsrelated to power generation equipment are included, with no unrelated business content.
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.