---
title: Charts, tables and infographics
category: read
summary: When the numbers in a Manifest deserve a picture, how to build a table, chart or infographic from them, and how to stay honest about what the data supports.
order: 17
updated: 2026-10-08
next: build-and-save-a-beacon
related: present-manifests-to-your-users, manifest-anatomy-field-guide
reference: /agents/personal-agent-onboarding.txt, /synorb-instructions.md
---
A good visual lets your user see in two seconds what a paragraph needs a minute to say. A bad visual invents precision. The method below lets the visual show only numbers that Synorb delivered.
## When to use this
Draw something when the story has numbers that **move or compare**:
- a change over time (a rate path, a price, a count rising day by day),
- a comparison across sources or entities (three banks' rate views side by side),
- a before and after (old pricing against new, last quarter against this one), or
- a result overview your user will scan (a run's hits by Stream, the window covered, allowance left).
Do not draw a comparison or a trend for numbers that do not compare. Three unrelated stories that each mention one figure are not a trend. A figure is still worth showing. Most weekly pulls have one value per `briefing.stories[].chart_series` entry, and that is the normal case.
For that case build a **figures panel**: 3 to 6 cards. Each card holds the value, its `unit`, the `qualifier` ("at least" for `at_least`), the `claim_text`, the item number and the source link. Take `parse_confidence: high` series first, one per story. Draw no bars, no ranking and no common scale across different units or measures.
Title the panel for what it is ("Figures from this week's items"), and label it "Created by your agent from this pull". Use a bar chart only when two or more items report the same measure in the same unit and period.
If your host cannot make images, say so in one line and give the same figures as a Markdown table.
Pick the richest form your display can show:
| Your display can | Use |
| --- | --- |
| Render HTML, SVG or images (an artifact, a dashboard, a web view) | An SVG chart, stat card or full infographic, as below |
| Run code (a notebook, a sandbox) | Any plotting library you have, saved as an image with alt text |
| Show only Markdown | A table, plus a text bar if it helps |
| Speak only (voice) | Lead with the one number that matters, in a sentence |
| Send email | An inline table and, if you attach an image, alt text and a link to the sources |
## Smallest working call
Pull Manifests. Full output (the default) carries all `facts` per Brief at `brief.body.facts`; compact output (`compact: true`) carries up to three at `brief.facts`. Each fact is a `label` and `value` pair traceable to the source:
```mcp billed
synorb-manifests
{"stream_ids": ["STREAM_ID"], "days": 14, "target_count": 10, "compact": true}
```
Now find the numbers that are truly comparable. This helper accepts a figure only if it is a plain number with an optional currency symbol or percent sign. A fact can also state its unit in a separate `unit` field and its time span in a `period` field. The helper keeps a label only if two or more sources report it in the same unit and for the same period. The [reference formatter](/agent-guides/present-manifests-to-your-users) applies the same rule in `comparison_table`, and also counts one article from several Streams as one source:
```python
import re
from html import escape
NUMBER = re.compile(r"^\s*([$€£]?)\s*(-?\d[\d,]*\.?\d*)\s*(%?)\s*$")
def parse_number(value):
match = NUMBER.match(str(value))
if not match:
return None
unit = match.group(1) or match.group(3)
return float(match.group(2).replace(",", "")), unit
def comparable_facts(manifests, min_sources=2):
by_label = {}
for m in manifests:
brief = m.get("brief") or {}
facts = brief.get("facts") or (brief.get("body") or {}).get("facts") or []
source = (m.get("source") or {}).get("source_name") or "Source"
for fact in facts:
parsed = parse_number(fact.get("value"))
stated = fact.get("unit") or "" # a fact can carry its unit apart from its value
if not parsed or not fact.get("label") or (parsed[1] and stated and parsed[1] != stated):
continue
by_label.setdefault(fact["label"], []).append(
{
"source": source,
"date": m.get("published_date"),
"value": parsed[0],
"unit": parsed[1] or stated,
"period": fact.get("period") or "",
"manifest_id": m.get("manifest_id"),
}
)
return {
label: rows
for label, rows in by_label.items()
if len(rows) >= min_sources
and len({r["unit"] for r in rows}) == 1
and len({r["period"] for r in rows}) == 1
}
```
A comparison table works on every display:
```python
def with_unit(value, unit):
number = format(value, ",g")
if unit in ("$", "€", "£"):
return unit + number
return number + unit if unit in ("", "%") else number + " " + unit
def markdown_table(label, rows):
period = rows[0].get("period")
lines = ["| Source | Date | " + label + (" (" + period + ")" if period else "") + " |", "| --- | --- | ---: |"]
for r in rows:
lines.append("| " + r["source"] + " | " + str(r["date"]) + " | " + with_unit(r["value"], r["unit"]) + " |")
return "\n".join(lines)
```
A bar chart is a self-contained SVG for any display that can render one. It has a title, a text description for screen readers, a value on every bar, and a source line:
```python
def svg_bars(title, rows, source_note=""):
if not rows or any(r["value"] < 0 for r in rows):
return None # use a table for negative values
width, left, bar_h, gap, top = 560, 190, 26, 10, 46
height = top + len(rows) * (bar_h + gap) + 34
peak = max(r["value"] for r in rows) or 1
desc = "; ".join(r["source"] + ": " + with_unit(r["value"], r["unit"]) for r in rows)
parts = [
'")
return "\n".join(parts)
```
Use them together:
```python
facts = comparable_facts(manifests)
for label, rows in facts.items():
print(markdown_table(label, rows))
svg = svg_bars(label, rows, "Sources: " + ", ".join(sorted({r["source"] for r in rows})))
```
## What you get back
For three sources that each report the same measure in percent, `markdown_table` returns:
```markdown
| Source | Date | Policy rate view |
| --- | --- | ---: |
| Bank A | 2026-10-02 | 4.25% |
| Bank B | 2026-10-03 | 4.5% |
| Bank C | 2026-10-05 | 4.75% |
```
`svg_bars` returns one `