# Odoo Demo CSV > Populate a connected Odoo demo database with data researched from a real company's website, or from a named industry. Use whenever a sales demo needs a database filled with realistic records. - Skill: `odoo-ps/odoo-demo-csv` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add odoo-ps/odoo-demo-csv` - Raw SKILL.md: https://api.skillmd.com/api/skills/odoo-ps/odoo-demo-csv/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Data & Analytics - Author: odoo-ps (https://skillmd.com/u/odoo-ps) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/odoo-ps/odoo-demo-csv --- # Odoo demo data A sales person names a company and its website — or just an industry — plus a dataset size. You research the target, write Odoo-import-ready CSVs, import them into the connected database, and verify the result: one continuous pass ending with a `SUMMARY:` line. Two words govern every step below. **Ground truth** — every record traces back to something you found on the target site: real product names, real prices, real photos. Where the site is silent, stay silent — price `0.00`, no image, and one line saying the data was thin. A demo built on invented products is the one failure this run cannot recover from. **Narrate** — one line before each step saying what you are about to do, one line after each `odoo-crud` call carrying its result ("Imported 25 partners", "Preview found 1 bad column, fixing"). The sales person watches this output and sees nothing else of the run. **No real people** — every customer, vendor and lead you invent is a fictional person or company; ground truth only covers the target company itself (its real products, prices, photos), never its staff's or its customers' contact details. Never let an invented contact resemble a real one, and never let its email or phone hint at any real company — including the target company itself: emails go on a domain that cannot resolve and carries no company name at all (`example.com` / `example.org` / `example.net` only — not `@-demo.test` or any variant that embeds a real name), phone numbers use a range reserved for fiction (e.g. `+1 555 0100`–`0199`, or a country's official drama/fake-number block) — never a real subscriber range. This applies to every `email`/`phone`/`mobile` column you write, in every model, with no exceptions — including rows for the target company's own res.company or res.partner record. ## The tool `odoo-crud [options]` reaches Odoo over XML-RPC. It is the only shell command available to you, and it reads its credentials from the environment. - **Blocking** — every call, `install-modules` included, returns once the work is done. Chain straight into the next command: there is nothing to wait for and nothing to poll. - Every call prints JSON `{"ok": true|false, ...}`. Read `ok` before moving on. - `odoo-crud --help` is the flag reference: the exact JSON shape of `--domain`, `--values`, `--args`, `--context`, `--filter` lives there. Read it rather than guess a shape. Use your own file tools to write the CSVs and your browsing tools to research. You have no shell beyond `odoo-crud` itself — writing CSV into the chat, or into a script you then try to run to generate the file, both go nowhere (the run to generate it is blocked); write the CSV content directly with your file tool. When a call fails, a record you know exists comes back missing, or a number reads zero after a green import, the cause is almost always one of the Odoo behaviours collected in [`ODOO-TRAPS.md`](ODOO-TRAPS.md) — read it then. ## Run ### 1. Research the target Browse the site and collect: the sector, the *real* products or services and their prices, the company's location (it drives country and currency), and the URLs of the logo and the product photos. Make each URL absolute — some sites emit protocol-relative `//cdn.example.com/…`, which needs an `https:` prefix. With no website given, generate coherent generic data for the stated industry instead. *Done when* you can name the products you will import and the country you will set on the company. ### 2. Probe the database This database's version decides which models, fields and apps exist — read them, never assume them. - `odoo-crud auth-check` — first command of the run. If it fails, stop and say why; nothing downstream can work. - `odoo-crud models --filter crm` — which models are there. - `odoo-crud fields product.template --filter 'name,list_price,barcode'` — one compact line per field (`many2one required -> res.partner`), including a selection field's allowed values. Filter by the columns you plan to write; dumping every field of `res.partner` or `product.template` buries the answer. - `odoo-crud install-modules crm stock sale_management account` — every module behind step 5's import order, in ONE call. Read the list straight off that order: `res.partner` → `contacts`, `product.template` → `product` and `sale_management`, stock on hand → `stock`, `crm.lead` → `crm`, plus `account` for invoicing. It installs *and* confirms before returning, and its JSON lists `not_installed` for anything that failed. *Done when* `install-modules` has come back with `not_installed` empty, and every column of every CSV you are about to write has appeared in a `fields` output, with its type and — for selection fields — its allowed values copied verbatim. ### 3. Configure the company Set country, company name and currency on `res.company` to the real location: find the id with `search-read`, then ``` odoo-crud write res.company --ids '[1]' --values '{"country_id": 241, "currency_id": 23}' ``` A currency that seems absent is archived, not missing — see the traps file. ### 4. Size the dataset The prompt gives **small**, **medium** or **big**. It is the scale of the whole business, so it applies to every model you populate. Anchor on the customer count and derive the rest in sector-adjusted proportion: | size | customers (anchor) | feel | |--------|--------------------|------------------------------------| | small | ~8 | a small shop — just enough to demo | | medium | ~25 | an established SMB | | big | ~80 | a large company, rich data | - **vendors** ≈ customers ÷ 6 — always a handful. - **products** ≈ a catalogue smaller than the customer base, fewer still for a services business. When the site has fewer genuine products than that, reach the number with real **variants** (sizes, flavours, formats) — ground truth outranks the target count. - **stock on hand** — a quantity for every storable product. - **leads** ≈ half to one× the customers. - **anything else you create** — sale orders, invoices, BoMs — scaled to match. Sizing one model richly and leaving its neighbours empty is what makes a demo look fake. ### 5. Generate and import Write each CSV to a file, then `odoo-crud import-csv --file `. Import in **dependency order** so every reference resolves: 1. **`res.partner`** — customers + vendors. Columns: `id` (e.g. `partner_client_1`), `name`, `is_company`, `street`, `city`, `email`, `phone`. 2. **`product.template`** — the flagship products you found. Columns: `id` (e.g. `product_1`), `name`, `list_price`, `standard_price`, `barcode`, plus this database's product-type and storability columns — whatever step 2's `fields product.template --filter 'type,storable'` showed, with the selection values verbatim. Whether storability is one of those values or its own boolean flag varies by version. 3. **stock on hand** — a *second pass over `product.template`*, one row per product you marked storable, carrying just `id` (the same external id as above) and `qty_available`. The ids already exist, so Odoo runs the same code path as typing a quantity on the product form: it creates the quant and applies it, and the goods really are on hand. 4. **`crm.lead`** — columns: `id` (e.g. `lead_1`), `name`, `partner_id/id` (an id from file 1), `expected_revenue`, `description`, `stage_id` (`New`, `Qualified` or `Proposition`). Write every `description` in the **customer's voice** — the two or three lines that prospect sent in, naming the real product they are asking about and what they need it for. Three rules bind every CSV: - **External IDs** — every row carries an `id` column and every reference is `field_id/id` pointing at one. This is what makes a re-import update the record instead of duplicating it, and it is non-negotiable. - **Quoting** — comma separator, and every text value wrapped in double quotes so `"Company, Inc."` stays one column. - **Language** — the site's language, or the one the sales person asked for. Asked for more than the default set (`mrp.bom`, employees, chart of accounts…)? Build those the same way: sized to the anchor, wired with external IDs, imported after whatever they reference. *Done when* every import has returned `ok: true` and its record count matches the number step 4 called for. On a failure, `ODOO-TRAPS.md` names the cause; fix the CSV and re-import — external ids make the retry safe. ### 6. Set images A product grid with real photos is what makes the demo land on screen. ``` odoo-crud set-image product.template --id --url odoo-crud set-image res.company --id 1 --url --field logo ``` Two details decide whether these calls land: - **Database id, not external id.** `--id` takes the integer Odoo assigned. Get them with one `search-read` on the model you just imported — it returns each record's `id` beside its `name`, which is what you match your photo URLs on. - **The field.** It writes `image_1920` by default; the company logo lives on `logo`, so that call carries `--field logo`. Records with no genuine image keep none — that is the only reason to skip one. *Done when* every product whose photo URL you collected in step 1 has had a `set-image` call return `ok: true`, and the logo is on the company. ### 7. Verify, then summarise Read back what you wrote — a green import is not proof: - `odoo-crud search-read product.template --fields '["name","qty_available"]'` — every storable product shows the quantity you imported. `qty_available` is the number that counts; `inventory_quantity` is not. - A `search-read` per model you populated, confirming the counts from step 4. *Done when* every model you touched has been read back and matches. Then print the `SUMMARY:` line.