aggregate
ChatGPTAggregate records server-side across deals, projects, ppas, financings, data_centers, or dg (US distributed-generation): pick a dataset and group by any 1-4 of its dimensions, returning per-cell count plus the dataset's magnitudes (MW, value, annual output, floor area; deals also carry median $/MW and n) in one call. Use for market-sizing, trend, and production-profile questions ("solar deals by country by year", "upcoming project capacity by commissioning year", "PPA MW by offtaker category"); the totals are computed server-side, without paginating raw rows. Rankings of entities (top buyers/owners/advisors) come from get_league_table, and the individual records behind a count from the search tools and render_exhibit.
export_results
ChatGPTExport a result set as a downloadable CSV, JSON, Markdown or Excel (xlsx) file - up to 500 rows, subject to the backend's per-query ceiling (the response says when fewer rows than requested were available) - and return a short-lived download link plus the row count. Use for requests to export, download, or get a spreadsheet / file of results, or for more rows than a search tool shows. Re-runs the same query the source tool runs, with the same filters, so the file reflects what that tool would report - the download re-runs the query when the link is opened, so it can differ if the data changes within the link's lifetime. Takes the same filters as search_ma_deals, search_projects, search_ppas, search_finance_deals, search_opportunities or search_data_centers (source "search"), get_league_table (source "league"), or get_deal_multiples (source "multiples"). "xlsx" suits a person opening it in Excel, "csv" other spreadsheet tools, "md" pasting a table into a document or chat, and "json" a script. The response states the row count, whether a cap applied, and when the link expires (15 minutes) - the facts that accompany the link. Returns metadata only - never the file contents.
find_advisor_records
ChatGPTFind renewable-energy records where a named advisor appears in any role - answers "which deals did a given bank or law firm advise on?" and returns the records themselves. One dataset per call: deals (default; M&A buyer/seller financial & legal, buyer technical, M&A financial; 6 roles), financings (financial/legal/technical), or ppas (offtaker/supplier advisor). Rankings of the most active advisors come from get_league_table. Optionally bound by announcement date (date_from/date_to), e.g. advisors active in 2021-2022. Substring match on the advisor field; case-insensitive and tolerant of punctuation/spacing variants (e.g. "JP Morgan" = "J.P. Morgan" = "JPMorgan").
find_comparables
ChatGPTFind records similar to a given base record across deals (default), ppas, or financings (dataset param) - answers "what deals are like this one?" for a specific deal, PPA or financing. Deals: sub-sector + state/region, falls back to sub-sector + country. PPAs/Financings: sector + country, falls back to sector-only. Magnitude +/-25%, date +/-2 years. Returns the individual similar records; aggregate median/mean $/MW and EV/EBITDA benchmarks over a comparable slice come from get_deal_multiples.
get_company_profile
ChatGPTGet a renewable energy company's consultant-grade profile by Company ID (company_id; alias id) or fuzzy name match (company_name; aliases company, name) - use for "tell me about company X" / company overview questions. Exactly one of the two is accepted per call; passing both returns an error. Composes identity, strategy, portfolio mix (by lifecycle/geography/sector), M&A activity (acquisitions, divestments, pipeline), financings, PPAs, executives, and branded source-links. Ambiguous names return candidates to disambiguate.
get_deal_multiples
ChatGPTCompute valuation benchmarks over an M&A deal slice. Returns statistics + the deals used (transparency). Use for 'what does this kind of deal trade for'. The individual comparable deals come from find_comparables; render_multiples shows these benchmarks as an interactive exhibit for requests to see them ("show me", "what does this trade for", "pull the comps as a table"), and with the same filters it shows this same slice. Two distinct $M/MW metrics come back in separate fields, each under its own name, and neither substitutes for the other: statistics.value_per_mw = implied enterprise value (EV) per MW of the transaction; statistics.developer_premium = the developer premium paid per MW for development-stage pipeline. developer_premium is a development-stage metric, NA for operational and in-construction assets (count 0 = NA, not a $0 value), which are benchmarked on value_per_mw. statistics.ebitda_multiple (EV/EBITDA) and statistics.revenue_multiple are the earnings-based multiples. For a non-Corporate development-stage slice (asset_lifecycle='Under Development' or a development_stage filter), developer_premium is the headline metric and value_per_mw / ebitda_multiple / revenue_multiple return null (pool count kept): enterprise and earnings multiples are noise for pre-operational assets. The $/MW metrics (value_per_mw, by_lifecycle, by_dev_stage, developer_premium) cover asset deals only: Corporate/platform deals, whose EV bundles pipeline, team and brand (so $/MW is meaningless), are excluded and reported in the corporate block (deal_count, ebitda_multiple, revenue_multiple, total value) on earnings multiples. Deal identity is deal_id: deals sharing MW, sector, country and value are not necessarily the same deal. A render_multiples card already on screen with the same benchmarks is complete.
get_deal_with_records
ChatGPTEnumerate the records linked to one M&A deal - its Projects, Financings, PPAs, and Opportunities - fetched by Deal ID. The deal's own story, structure, rationale, or portfolio split is in the deal narrative from get_record_detail (dataset "deals") - that analyst summary is the authoritative source and is returned by default. Use this tool when that narrative is absent/empty, or for listing the linked records ("and its projects" / "with PPAs" / "everything tied to the deal"). For BESS, energy capacity (MWh) and duration (= MWh / MW) are stated in deal_heading / the summary Asset Details, not a structured field; storage_mwh and storage_duration_hours are extracted best-effort from the heading and are null when it states no MWh.
get_league_table
ChatGPTRank entities (buyers/sellers/advisors, owners/developers/contractors, offtakers/suppliers, financiers, etc.) over a slice of one of six datasets (dataset = deals | projects | ppas | financings | opportunities | dg; default deals) by record_count, total_mw, total_value_musd, or total_output_gwh. dataset=dg ranks US distributed-generation (rooftop/community solar + storage, US-only) owners/developers by record_count or total_mw. Use for "top N buyers" / "most active developers" / "rank lenders by amount" queries. render_league displays the ranking as an interactive exhibit - a ranked bar chart and, where records are available, the underlying records those entities appear on - for requests to see the rankings themselves on any dataset, dg included ("show me the top", "who is most active", "rank the lenders", "league table"); a typed Markdown table of the rankings is static by comparison. The dimension to rank is entity_type (aliases entity, dimension, role; precedence entity_type > entity > dimension > role); its valid values depend on dataset. An entity_type or metric not valid for the chosen dataset returns a clean error, not a crash. Given the same filters as this query, the render_league card shows the same slice. When a card already on screen carries the same rankings, a second one only repeats what the user is already looking at.
get_record_detail
ChatGPTFetch one record by its business ID from a chosen dataset (deals, projects, financings, ppas, opportunities). Direct fetch (no search filters) - use for the full detail and analyst narrative of one deal, project, financing, PPA or opportunity whose ID is known; a deal's linked records are listed in full by get_deal_with_records. It is the deal-detail path (it replaced the former standalone get_deal_detail tool); for dataset=deals, include_discrete_advisors=true also surfaces the 6-way discrete advisor breakdown (buyer financial, legal and technical; seller financial and legal; M&A financial) and the numeric record_id. A deal value is the consideration for that transaction; an absent value means "not disclosed", not $0. A figure quoted in the heading or summary narrative may describe a whole portfolio, pipeline, or company valuation rather than this deal; the structured value is the authoritative deal value. For projects, optionally include FERC PPA pricing. The per-dataset analyst summary (the record's own narrative) is returned by default (include_summary defaults true) and is the authoritative source for a record's story, structure, or portfolio split, ahead of any linked record; include_summary=false trims it. For dataset=deals the response also lists linked-record reference IDs (linked_project_ids, linked_ppa_ids, linked_financing_ids, linked_opportunity_ids) plus a linked_record_counts summary; these are references only; get_deal_with_records hydrates them into full project/PPA/financing/opportunity rows. For BESS, energy capacity (MWh) and duration (= MWh / MW) are stated in deal_heading / the summary Asset Details, not a structured field; storage_mwh and storage_duration_hours are extracted best-effort from the heading and are null when it states no MWh. Deal identity is deal_id: two deals that share the same MW, sector, country, and value can be distinct deals, so matching attributes do not identify a duplicate. Pass dataset="dg" with a DG Project ID (alias dg_project_id) to fetch ONE US distributed-generation project (rooftop / community solar + storage). DG is a separate Postgres dataset: it returns its own 17 fields, carries NO source link, and none of the deals-specific options (include_ferc, include_summary, include_discrete_advisors, linked-record IDs) apply to it.
open_report
ChatGPTOpen one specific Enerdatics Tier-1 research report (category='Report') to read in-chat, page by page, with a link to the full PDF. Use for 'show me / open / read / walk me through the report' requests to view a report's pages - the widget renders on every call. Discussing, searching inside, or summarising a report is served by search_report, which returns cited excerpts, not a rendered reader. slug or report_id opens a specific report. With neither, it opens the most recent report by publication date ("the latest report", "the H1 report") with no identifier or prior lookup in another tool. Not a deal-data source: questions like 'which deals happened in Spain', 'biggest BESS transactions', or 'valuation multiples' are answered by search_ma_deals / get_deal_multiples / get_league_table from the live deal database; a report PDF is a point-in-time document that cannot authoritatively answer a live data question. Some source documents are flattened page images or otherwise unreadable as text; for those, pages comes back empty and notice says the text could not be extracted (a correct, non-error outcome, never a silent empty result), and the full-PDF link is still returned. pages, the document's own extracted text, is the answer source: when it is non-empty the report has genuinely been read and can be summarised, quoted and cited like any document. page_images is a separate channel: rendered pages the user sees in the widget, whose images go straight to the display and are not part of the model's context. Its presence says nothing about whether the report is readable; that is decided by pages alone, and a report can carry both full text and page images. An empty pages means the document has not been read, whatever page_images shows the user. A report that does not exist gets a normal, non-error response naming what was looked for, not a tool failure.
render_exhibit
ChatGPTDisplay search results as an interactive table the user can sort, filter and page through, over deals, projects, ppas, financings, opportunities, data_centers or dg - for requests where the rows themselves are the answer: "show me", "list", "pull the comps", "give me N deals/projects/PPAs", or any request expected to read as a table. Entity rankings (top buyers, most active advisors, biggest lenders) come from get_league_table and render_league, even when phrased as "list" or "show me". With no filters, a call returns only the first 50 rows of the whole dataset (the 50 most recent, except on data_centers and dg) and the response states the full total (dg excepted); sector, geography (country, region, power market), period (dates) and, for deals, deal type narrow it to a specific market, so an unfiltered request ("show me some deals") is an arbitrary slice of the most recent records rather than an answer about any particular market. It is the display step for search results, not a second copy of the data: re-fetching rows a search already returned is expected and cheap (the usual shape: a search to explore or verify, then this once to show the result). A Markdown table typed into a reply is static, capped by what can be retyped, and loses the filter chips and the link to every record. Filters follow the chosen dataset and match its search tool; a filter the dataset does not support is refused by name rather than quietly ignored. Several sectors ("solar and BESS") or lifecycles fit one table and one call (sectors / lifecycles). Shows up to 50 rows with the true total (export_results writes up to 500 rows to a file). Rows arrive newest first unless sort_by "mw" or "value" puts the largest first, so a biggest/largest question gets the 50 largest rather than the 50 most recent. The caption states the order actually used; an unsorted table is not a ranking.
render_ferc_pricing
ChatGPTDisplay FERC PPA pricing as an interactive exhibit: a generation-weighted average price chart (split by sector, power market, project, offtaker type or offtaker - sector by default) plus the underlying priced records in a filterable table. It is the display step after search_ferc_pricing, for answers that read best as a chart or a browsable table. offtake_category, price_type, power_market, sector, customer_company and project_name filters all narrow both the chart and the table. customer_company scopes the card to a named buyer (e.g. Google in MISO) and project_name to a single project; this card answers both, rather than a raw record search. Scope is US FERC PPA pricing from 2023 onward.
render_league
ChatGPTDisplay a league-table ranking as an interactive exhibit: a ranked bar chart and, where records are available, the underlying records those entities appear on (unlike the raw ranking), sortable and filterable. It is the display step for ranking requests ("show me the top", "who is most active", "rank the lenders", "league table"), the way render_multiples is the display step for "what is it worth". Runs the identical engine get_league_table uses, over one of six datasets (deals, projects, ppas, financings, opportunities, dg), so a bar can never disagree with a ranking already quoted from get_league_table. entity_types (up to 4) ranks several roles together in one card as separate groups - e.g. a side-less "top financial advisors" is ["buyer_financial_advisor","seller_financial_advisor","ma_financial_advisor"] - where one call per role would render separate cards. countries or states_us (one or the other, not both) ranks a multi-value geo slice as one card. It renders the ranking get_league_table returns as the presentation surface, not a second copy of it, so a ranking already quoted from get_league_table can be rendered here. The underlying records are fetched fresh from the Enerdatics database for deals/projects/ppas/financings parent roles; a role whose records are not available here (a financings child role - lender/investor/underwriter - or dataset=opportunities/dg) still renders its ranking, with a plain statement that records are unavailable for it rather than an empty table presented as fact.
render_multiples
ChatGPTDisplay valuation benchmarks as an interactive exhibit: a KPI row, a chart, and the deal table underneath, all filtering together. It is the display step for pricing questions: the ranked numbers with the deals behind them, rather than a figure typed into prose. The individual comparable deals, as opposed to a benchmark over them, come from find_comparables. Use it when what a deal or asset class is worth is the answer - "valuations", "what does this trade for", "what did X trade at", "benchmark", "multiples", "comps pricing", "$/MW", "EV/MW", "how much per MW", "developer premium". A retyped Markdown table is static, loses the source links and cannot be filtered. It renders what get_deal_multiples computes as the presentation surface, not a second copy of the data, so re-fetching a slice already computed there is expected and cheap. The usual shape is get_deal_multiples to explore or verify, then this once to show the result. Takes the same filters as get_deal_multiples (country, region, sector, segment, MW, dates, lifecycle, development stage, power market, and more) and computes them with the same engine, so the figures match what get_deal_multiples returns for the identical slice. Chart types (the chart param): candle, line, bubble, bar; omitted, the tool picks from the slice's shape. A deal multiple needs a disclosed value, so deals that never disclosed one are excluded from this exhibit entirely - not from the median only, but from the KPI row, the chart and the table alike. Every count on the card is therefore a count of priced deals. When a slice has none, the tool says so instead of drawing an empty benchmark. Nothing is ever shown as a $0 price.
search_data_centers
ChatGPTSearch data centers (Beta) by country, state/region, county, owner, name, type, size, lifecycle, power market, power source, PPA-signed, MW capacity, area, and year built. Returns data center records with owner, capacity, power source, PPA status, and source-links. render_exhibit with dataset:"data_centers" displays these data centre facilities as an interactive sortable table with source links - the display for requests to see the rows themselves ("show me", "list", "pull the sites"), where a typed Markdown table has neither sorting nor links. It takes the same filters as this search and shows the same slice when given them; a filter it has no typed param for is accepted in filters[], which is validated against that dataset and refused by name, never silently dropped. A card already on screen with the same rows is complete; a second render_exhibit call for them repeats that table. (Beta) - this table is newer and may be less fully populated than the M&A/PPA tools.
search_dg_projects
ChatGPTSearch US-only distributed-generation projects: small-scale (rooftop / C&I / community / sub-utility-scale) solar and battery systems across US states and ISO/RTO power markets. Filterable by sector, sub-sector, installation type, customer type, power market, MW capacity, lifecycle, US state/region, county, developer, ownership, and operational dates. This dataset contains no non-US projects and has no country filter (US is implicit); utility-scale and non-US projects are in search_projects. render_exhibit with dataset:"dg" displays these rows as an interactive sortable table - the display for requests to see the rows themselves ("show me", "list", "pull the community solar projects"), where a typed Markdown table cannot be sorted. It takes the same filters as this search and shows the same slice when given them. That card shows the first 50 matches and carries no total match count, so the number of matches beyond those 50 is unknown. A card already on screen with the same rows is complete; a second render_exhibit call for them repeats that table. Returns lean project records.
search_ferc_pricing
ChatGPTSearch US FERC-reported settled PPA pricing (weighted-average $/MWh per period) by ISO/power market (PJM, CAISO, MISO, SPP, ISO-NE, NYISO), state, sector, off-taker, project ownership, customer company, or utility service area. This is the indicative measure of the price a US project realizes. ERCOT has no FERC-reported rows, so power_market="ERCOT" returns none; Texas projects in SPP or MISO are covered and are reached with state_region="Texas". A market with no rows has no FERC price here; rows from another ISO describe a different market. Off-taker is a multi-value list of every party named on the contract, so one contract counts under each named party; customer company is the single counterparty that settled the transaction, so grouping, ranking or averaging by customer company counts each transaction once. Utility service area holds a balancing-authority or hub code (MISO, AMIL, EES, TVA), not a utility name. Global PPA contract terms (supplier, duration, on/offsite) come from search_ppas. render_ferc_pricing displays these priced rows as an interactive exhibit - a realized-price chart over time plus the underlying contract rows with source links - for requests to see the rows themselves ("show me", "list", "pull the FERC prices", "chart the realized price"); it is a richer surface than the same rows typed out as a Markdown table. It accepts the same filters, so the same filters render the same slice. When a card already on screen carries the same priced rows, a second call adds a duplicate table of what the user is already looking at.
search_finance_deals
ChatGPTSearch renewable energy financing transactions (debt + equity raises) by company, country, sector, type, amount, co-located storage, power market, and dates. Company and asset acquisitions (M&A) are in search_ma_deals. Returns financings with amount breakdown and source-links. Rankings of the companies, sponsors, investors, lenders, underwriters and advisors behind these raises ("top financiers", "most active lenders") come from get_league_table with dataset=financings. render_exhibit with dataset:"financings" displays these debt and equity raises as an interactive sortable table with source links - the display for requests to see the rows themselves ("show me", "list the raises", "who raised what"), where a typed Markdown table has neither sorting nor links. It takes the same filters as this search and shows the same slice when given them; a filter it has no typed param for is accepted in filters[], which is validated against that dataset and refused by name, never silently dropped. A card already on screen with the same rows is complete; a second render_exhibit call for them repeats that table. For "Solar+BESS" / "solar plus storage" / "hybrid" / "co-located" financings, the filter is sector=<generation sector, e.g. Solar> + co_located_storage=Yes; the compound is not a sector value, and sector=Battery means a standalone battery. Financings carry a power market / ISO tag, filterable with power_market (ERCOT, CAISO, PJM, ...); some are Undisclosed on power market. filters[] accepts any known financing field without a typed param (e.g. financing_sub_type, amount_tax_equity_musd); an unknown field is rejected with the valid list, never silently ignored.
search_insights
ChatGPTList and search Enerdatics research insights (analysis notes, reports, executive briefings) by sector, category, country, continent, publication date, or free-text query - answers "what has Enerdatics published on X?" with summaries and links. With no filters, returns the most recently published insights. Returns each insight with its summary, sector, category, PDF link, and source-link. Country/continent narrow the results but are not echoed back in each record; the PDF/source link shows each insight's geography.
search_ma_deals
ChatGPTSearch renewable energy M&A deals by country, region (continent or market grouping - e.g. Europe, APAC, LATAM), sector, segment, MW, date (announced or completed), deal type, deal status, deal value, equity transacted, asset lifecycle, development stage, US state, power market, buyer/seller name, and buyer/seller classification. Returns deal records with buyer, seller, MW, value, and source-links for verification. Requests naming a different subject - advisors, opportunities, projects, PPAs, most-active buyers - belong to those tools even when phrased as "list" or "show me". Best when the deal rows are an input: to compare, count, rank, reason over, or feed another tool such as get_record_detail or find_comparables. Safe to call several times while working toward an answer; with no filter, a call returns up to three narrowing questions (country/region, sector/segment, time window/status) instead of rows. render_exhibit is the display for requests to see the rows ("show me", "list", "pull the comps"): an interactive sortable table with source links. A card already on screen with the same rows is complete. One render_exhibit call with sectors:["Solar","Battery"] puts several sectors in one table; one call per sector gives separate tables. Coverage: thousands of deals across every country, sector and year. Notes: - All enum values are case-sensitive platform labels, matched verbatim. - BESS MWh and duration (MWh / MW) are in deal_heading / the summary Asset Details, not a structured field; storage_mwh and storage_duration_hours are best-effort extracts from the heading, null when no MWh is stated. - Deal identity is deal_id: matching MW, sector, country and value do not make two deals the same. - stealth_only=true: "stealth deals" / "deals Enerdatics sourced" / "proprietary deals"; rows carry stealth_sourced and stealth_note (the citable reason a deal is stealth).
search_opportunities
ChatGPTSearch renewable energy investment opportunities and deal pipeline by country, continent, US state, sector, status, category, deal type, stage, lifecycle, power market, MW, deal value, and announced date. Returns open-pipeline opportunities with acquisition target, advisor, deal value, and source-links. Only open-pipeline opportunities (status Confirmed or Rumoured) are returned; cancelled or closed opportunities are out of scope, and announced or completed M&A transactions are in search_ma_deals. render_exhibit with dataset:"opportunities" displays these currently-for-sale assets as an interactive sortable table with source links - the display for requests to see the rows themselves ("show me", "list", "what is on the market"), where a typed Markdown table has neither sorting nor links. It takes the same filters as this search and shows the same slice when given them; a filter it has no typed param for is accepted in filters[], which is validated against that dataset and refused by name, never silently dropped. A card already on screen with the same rows is complete; a second render_exhibit call for them repeats that table.
search_ppas
ChatGPTSearch renewable energy Power Purchase Agreements (PPAs) by country, sector, off-taker, supplier, off-taker category, onsite/offsite, power market, MW capacity, duration, and announcement date. Returns PPA records with off-takers, suppliers, tariff ($/MWh), and source-links. Indicative US settled $/MWh per quarter, a different measure from the contract tariff here, comes from search_ferc_pricing. render_exhibit with dataset:"ppas" displays these PPA rows as an interactive sortable table with source links - the display for requests to see the rows themselves ("show me", "list", "pull the PPAs"), where a typed Markdown table has neither sorting nor links. It takes the same filters as this search and shows the same slice when given them; a filter it has no typed param for is accepted in filters[], which is validated against that dataset and refused by name, never silently dropped. A card already on screen with the same rows is complete; a second render_exhibit call for them repeats that table. For "Solar+BESS" / "solar plus storage" / "hybrid" PPAs, the filter is sector=<generation sector, e.g. Solar> - co-located storage is not a PPA search filter (the PPA co-located field is not server-side queryable); a PPA's co-located status is shown by get_record_detail(dataset=ppas). US PPAs are tagged by power market / ISO: power_market="MISO" (or PJM, CAISO, ERCOT, SPP, ISO-NE, NYISO) filters PPA contracts by grid. filters[] accepts any known PPA field without a typed param (e.g. contract_type, virtual_ppa); an unknown field is rejected with the valid list, never silently ignored.
search_projects
ChatGPTSearch renewable energy projects by country, sector, solar technology (PV/CSP), power market (ISO/RTO), route-to-market, co-located storage, MW capacity, lifecycle stage, and dates. Returns project records with capacity, investment, and source-links. Covers utility-scale projects worldwide; US small-scale distributed generation (rooftop / C&I / community / behind-the-meter solar) is in search_dg_projects. The US renewable interconnection queue is covered by power_market (the ISO/RTO, e.g. PJM/ERCOT/CAISO) and queue_status (the Grid connection status ladder, e.g. "IA Executed", "Withdrawn"), with has_interconnection_id=true selecting projects with a known queue position; the aggregate tool groups by power_market and/or queue_status to size the queue. render_exhibit with dataset:"projects" displays these utility-scale generation and storage assets as an interactive sortable table with source links - the display for requests to see the rows themselves ("show me", "list", "pull the projects"), where a typed Markdown table has neither sorting nor links. It takes the same filters as this search and shows the same slice when given them; a filter it has no typed param for is accepted in filters[], which is validated against that dataset and refused by name, never silently dropped. A card already on screen with the same rows is complete; a second render_exhibit call for them repeats that table. For "Solar+BESS" / "solar plus storage" / "wind+battery" / "hybrid" / "co-located" projects, the filter is sector=<generation sector, e.g. Solar> + co_located_storage=Yes (the authoritative hard co-located filter); the compound is not a sector value, and sector=Battery means a standalone battery. text_search does a free-text substring match on the project name.
search_report
ChatGPTSearch inside Enerdatics' published Tier-1 research reports (PDF studies, category='Report') for what they say about a topic - returns cited excerpts (title, slug, page number) you can quote. Every report in the library is findable by title/summary even for the few whose source PDF could not be read; those return a clearly-labeled metadata match, never a fabricated page quote, and the response always states how many reports have readable body text versus none. Use for 'what does our research say about X' / 'find a report on Y'. Not a deal-data source - questions about specific transactions, valuations, deal multiples, or league tables (e.g. 'which deals happened in Spain', 'biggest BESS deals', 'valuation multiples') are answered by search_ma_deals / get_deal_multiples / get_league_table from the live deal database rather than from a report PDF.
submit_feedback
ChatGPTSubmit free-text feedback from the user to the Enerdatics team, who review every message. Use when something is confusing, wrong, slow, missing, or worth keeping.
aggregate
Claudeexport_results
Claudefind_advisor_records
Claudefind_comparables
Claudeget_company_profile
Claudeget_deal_multiples
Claudeget_deal_with_records
Claudeget_league_table
Claudeget_record_detail
Claudeopen_report
Clauderender_exhibit
Clauderender_ferc_pricing
Clauderender_league
Clauderender_multiples
Claudesearch_data_centers
Claudesearch_dg_projects
Claudesearch_ferc_pricing
Claudesearch_finance_deals
Claudesearch_insights
Claudesearch_ma_deals
Claudesearch_opportunities
Claudesearch_ppas
Claudesearch_projects
Claudesearch_report
Claudesubmit_feedback
Claude