Dcipher Analytics

Create a News Knowledge Base by scanning news articles — use this for horizon scanning, trend monitoring, and industry news. Takes ready Opoint query params, but there is no keyword-suggestion tool: derive concrete search keywords yourself from the topic the user gave, and state the ones you chose so they can correct them. Aim for roughly 5,000-15,000 articles in the resulting KB — steer queries, filters, and time windows toward that range. Omit `schedule` for a ONE-TIME fetch; pass `schedule` to stand up a self-refreshing flow instead, where the first run fires almost immediately (now+delay) and each subsequent run fetches fresh and creates a NEW timestamp-suffixed KB. Returns the flow id immediately; runs are asynchronous and materialize the KB on completion. For a ONE-TIME fetch (no schedule), call wait_for_flow next with the returned flowId and keep calling it until done — do not ask the user whether to wait or check status, and do not hand back control in between. For a scheduled flow, this call is enough — confirm the schedule and hand back control without waiting on the first run. get_last_flow_run_status is only for a one-off check that should not block, such as the latest run of a scheduled flow. Pass insightBoosterId to auto-link each KB to that project on completion; omit it to attach later with attach_knowledge_base_to_insight_booster. If you chose the data source or its key parameters yourself rather than the user specifying them, state them and confirm with the user before calling.

create_news_kb

Create a Knowledge Base that GENERATES its data with a research agent scanning public web sources — use this for desk research such as government reports, academic publications, and filings. You define the agent task and (optionally) placeholder variables; the agent runs once per variable combination, writing the default output columns (Text, Source, URL, Date) plus the input variables. Omit `schedule` for a ONE-TIME run; pass `schedule` to re-run the task on a cadence, where each run creates a NEW timestamp-suffixed KB. Returns the flow id immediately; runs are asynchronous and materialize the KB on completion. For a ONE-TIME run (no schedule), call wait_for_flow next with the returned flowId and keep calling it until done — do not ask the user whether to wait or check status, and do not hand back control in between. For a scheduled flow, this call is enough — confirm the schedule and hand back control without waiting on the first run. get_last_flow_run_status is only for a one-off check that should not block, such as the latest run of a scheduled flow. Pass insightBoosterId to auto-link each KB to that project on completion; omit it to attach later with attach_knowledge_base_to_insight_booster. If you chose the data source or its key parameters yourself rather than the user specifying them, state them and confirm with the user before calling.

create_research_agent_kb

Create a Social-media Knowledge Base by scanning social posts — use this for voice-of-the-customer, sentiment, and community discussions. Takes ready Netfeedr query params and per-platform post limits, but there is no keyword-suggestion tool: derive concrete search queries yourself from the topic the user gave, and state the ones you chose so they can correct them. Aim for roughly 5,000-15,000 posts in the resulting KB — steer queries, filters, and time windows toward that range. Omit `schedule` for a ONE-TIME fetch; pass `schedule` to stand up a self-refreshing flow instead, where the first run fires almost immediately (now+delay) and each subsequent run fetches fresh and creates a NEW timestamp-suffixed KB. Returns the flow id immediately; runs are asynchronous and materialize the KB on completion. For a ONE-TIME fetch (no schedule), call wait_for_flow next with the returned flowId and keep calling it until done — do not ask the user whether to wait or check status, and do not hand back control in between. For a scheduled flow, this call is enough — confirm the schedule and hand back control without waiting on the first run. get_last_flow_run_status is only for a one-off check that should not block, such as the latest run of a scheduled flow. Pass insightBoosterId to auto-link each KB to that project on completion; omit it to attach later with attach_knowledge_base_to_insight_booster. If you chose the data source or its key parameters yourself rather than the user specifying them, state them and confirm with the user before calling.

create_social_kb

Delete a single filter from an Insight Booster project, identified by filter id and filter type.

delete_insight_booster_filter

Delete an AI-generated report template. Reports already generated from it are not affected.

delete_insight_booster_report_template

Delete a view from an Insight Booster project.

delete_insight_booster_view

Delete a workbench from an Insight Booster project.

delete_insight_booster_workbench

Detach every Knowledge Base from an Insight Booster project. Irreversible without re-attaching.

detach_all_knowledge_bases_from_insight_booster

Detach a single Knowledge Base from an Insight Booster project.

detach_knowledge_base_from_insight_booster

Return the union of metadata schema entries from all KBs attached to the project. Useful to drive filter UIs and validation.

fetch_project_metadata_schema

Return the last N messages from the research-bot conversation

fetch_research_chatbot_conversation_history

Generate a new AI-generated report on a workbench by applying a saved report template. Returns the new report id.

generate_insight_booster_workbench_report_from_template

Generate structured report sections for a given interest area. The returned object contains sections suitable for inclusion in a generated report. This operation can be useful to decide on Report Generator sections based on project data.

generate_report_sections_from_landscape

MCP version of generate_bump. Returns {'result': ...}. Projection is applied to the value inside result, not to the result envelope. Before projection, result has the BumpChartResponse schema. Top-level fields are day, week, month, quarter, semiyear, and year. Each top-level field is an array of BumpChartDataPoint objects. Each data point has x: string (time bucket label), y: string (series or category label), and val: integer (rank or numeric value). Always provide projection as {'pipeline': [...]} using MongoDB aggregation syntax. Use $project for selecting granularities such as week or month. Use $map when reshaping data point arrays. Use $slice to limit large arrays. Use $filter to keep only data points matching x, y, val, or other conditions. Do not project result.* paths; project day, week, month, quarter, semiyear, year, and their nested fields directly. Use the smallest projection that returns only the fields needed to answer the user, because the projected result will be sent to an LLM.

get_bump_result

Resolve an uploaded file by name. Returns `{ ready: true, file }` once the file is registered in storage, or `{ ready: false }` while it is not yet available. Use this to poll after the file was uploaded through the app — file registration happens asynchronously once the bytes land, so a just-uploaded file may briefly report not-ready before it can be used.

get_file_by_name

MCP version of analyze_growth. Returns {'result': ...}. Projection is applied to the value inside result, not to the result envelope. Before projection, result is an array of GrowthChartTopicPoint objects. Each topic point has id: string, title: string, summary: string, examples: array, metadata: object, size: integer, growths: object, and volumes: object. examples is an array of source example objects. Each example has text: string, source: string or null, and url: string or null. metadata is an object with string keys and string/number values. growths is an object keyed by time-bucket label with number or null values. volumes is an object keyed by time-bucket label with integer values. Always provide projection as {'pipeline': [...]} using MongoDB aggregation syntax. Use $project for field selection and reshaping. Use $map when selecting fields from nested examples. Use $slice or $limit to reduce large arrays. Use $filter or $match to keep only topics matching title, size, growth, volume, or metadata conditions. Do not project result.* paths; project id, title, summary, examples, metadata, size, growths, volumes, and their nested fields directly. Use the smallest projection that returns only the fields needed to answer the user, because the projected result will be sent to an LLM.

get_growth_result

Fetch full details for an Insight Booster project by id, including workbenches, attached Knowledge Bases, filters, and research-bot configuration. Use this before making changes to inspect current state.

get_insight_booster

Get the research-bot configuration for an Insight Booster project (allowed domains, rate limit, UI prefs).

get_insight_booster_research_bot_config

Get a single workbench by id along with its current AI-generated report output.

get_insight_booster_workbench

Return aggregated statistics for the requested metadata field (categoric counts, numeric min/max/mean, or date range).

get_kb_field_statistics

Retrieve the maximum broadness level supported for `analyze_growth`. This value determines the highest level of topic aggregation available for the landscape. Use to validate or bound broadness parameters in growth analysis requests.

get_landscape_max_broadness

MCP version of generate_landscape. Returns {'result': ...}. Projection is applied to the value inside result, not to the result envelope. Before projection, result has the LandscapeGeneratorResponse schema. Top-level fields are documents and topics. documents is an array of document point objects. Each document has id: string (document identifier), x: number and y: number (map coordinates), radius: number or null (visual size), and color_value: string or null (visual color bucket). topics is an array of topic level objects. Each topic level has level: integer (broadness level) and topic_points: array. topic_points is an array of topic objects. Each topic has id: string, x: number, y: number, radius: integer, title: string, summary: string, examples: array, and metadata: object with string keys and string/number values. examples is an array of source example objects. Each example has text: string, source: string or null, and url: string or null. Always provide projection as {'pipeline': [...]} using MongoDB aggregation syntax. Use $project for top-level field selection and reshaping. Use $map when selecting fields from nested arrays such as topics, topic_points, or examples. Use $slice to limit large arrays. Use $filter to keep only array items matching a condition. Do not project result.* paths; project documents, topics, and their nested fields directly. Use the smallest projection that returns only the fields needed to answer the user, because the projected result will be sent to an LLM.

get_landscape_result

Get the LATEST run status of a KB flow, without blocking — works for both one-shot and scheduled flows. After a one-shot create (create_news_kb / create_social_kb / create_research_agent_kb), use wait_for_flow instead; reach for this one when a single non-blocking check is all you need. It returns the task status and, once complete, the resolved knowledgeBaseId to attach downstream. For a scheduled flow it reports the MOST RECENT run on the schedule cadence — mind the run times: between runs it shows the last completed run, and the returned knowledgeBaseId is the first run's KB (each run materializes a new timestamp-suffixed KB). When a run FAILED, the returned message carries the failure reason; it is omitted otherwise.

get_last_flow_run_status

MCP version of generate_matrix. Returns {'result': ...}. Projection is applied to the value inside result, not to the result envelope. Before projection, result is an array of GenerateMatrixUnitResponse objects. Each matrix cell has id: integer, row: string, column: string, output: string, size: integer, and references: array. references is an array of objects with source: string and url: string or null. Always provide projection as {'pipeline': [...]} using MongoDB aggregation syntax. Use $project for field selection and reshaping. Use $map when selecting fields from nested references. Use $slice or $limit to reduce large arrays. Use $filter to keep only cells matching row, column, size, or other conditions. Do not project result.* paths; project id, row, column, output, size, references, and their nested fields directly. Use the smallest projection that returns only the fields needed to answer the user, because the projected result will be sent to an LLM.

get_matrix_result

Compute and return the earliest and latest timestamps available across the project's knowledge bases.

get_project_time_range

MCP version of generate_radar. Returns {'result': ...}. Projection is applied to the value inside result, not to the result envelope. Before projection, result is an array of RadarSegment objects. Each segment has id: string, title: string, and bubbles: array. bubbles is an array of RadarBubble objects. Each bubble has id: string, title: string, summary: string, count: integer, angularScale: number, radialScale: number, sizeScale: number or object, references: array, examples: array of strings, and isPredefined: boolean. references is an array of source example objects. Each reference has text: string, source: string or null, and url: string or null. Always provide projection as {'pipeline': [...]} using MongoDB aggregation syntax. Use $project for field selection and reshaping. Use $map when selecting fields from nested arrays such as bubbles or references. Use $slice or $limit to reduce large arrays. Use $filter to keep only segments or bubbles matching a condition. Do not project result.* paths; project id, title, bubbles, and their nested fields directly. Use the smallest projection that returns only the fields needed to answer the user, because the projected result will be sent to an LLM.

get_radar_result

MCP JSON version of generate_report. Returns {'result': ...}. Projection is applied to the value inside result, not to the result envelope. Unlike the REST endpoint, this endpoint returns generated report section JSON instead of a streamed file. Before projection, result is an array of section body objects. Each section body has output: string (generated section text), references: array, and match_score: integer (reference match count; 0 for agentic-generated sections). references is an array of source reference objects; each reference has source: string or null and url: string or null. Always provide projection as {'pipeline': [...]} using MongoDB aggregation syntax. Use $project for field selection and reshaping. Use $map when selecting fields from nested references. Use $slice or $limit to reduce large arrays. Do not project result.* paths; project output, references, match_score, and their nested fields directly. Use the smallest projection that returns only the fields needed to answer the user, because the projected result will be sent to an LLM.

get_report_result

Maps an open-ended interest area onto experimentally validated search keywords by calling the workflow backend search-query agent, which tries candidate queries against live News/social-media APIs instead of guessing keywords. Call this before creating a News or social-media knowledge base whenever the user has not already supplied concrete search keywords.

get_search_queries_from_agent

Perform a full-text or metadata-restricted search inside the specified knowledge base. This endpoint returns full texts.

kb_search_documents

List the caller's archived Insight Booster projects (soft-deleted). Use this when the user asks about previously archived projects or wants to restore one.

list_archived_insight_boosters

List the files stored in the caller’s organization, each with its name, group, extension and size. Use it to discover the exact stored name to pass to file-consuming tools.

list_files

List the Knowledge Bases attached to an Insight Booster project, with their current sync status.

list_insight_booster_knowledge_bases

List all AI-generated report templates in the caller's organization (compact form).

list_insight_booster_report_templates

List the caller's Insight Booster projects in compact form. Use this to discover what projects exist before fetching details on a specific one.

list_insight_boosters

Knowledge Bases are structured text datasets indexed semantically, supporting querying, filtering, and RAG; once ready, one can be linked to Insight Booster projects. List the Knowledge Bases in the caller's organization. Use this to recover a knowledgeBaseId you no longer have, or to discover what Knowledge Bases exist before attaching one to an Insight Booster or inspecting it — do NOT guess an id. Each entry includes the id, name, status, document volume, creation/update timestamps, whether it is fed by a scheduled flow, and the Insight Booster projects it is linked to. Optionally pass nameContains to narrow the list by a case-insensitive name substring.

list_knowledge_bases

List AI-generated report templates owned by the caller (subset of list_insight_booster_report_templates).

list_owned_insight_booster_report_templates

Rename an Insight Booster project.

rename_insight_booster

Rename a workbench within an Insight Booster project.

rename_insight_booster_workbench

Reorder workbenches within an Insight Booster project. Provide the desired final position list.

reorder_insight_booster_workbenches

Restore a previously archived Insight Booster project so it appears in the active project list again.

restore_insight_booster