bug: Use repr() for chat tool calls when exporting markdown

[schloerke]

If we're putting the value in a python code block, then it is safe to assume the code can be run. repr() should return eval-able objects (which do not need an extra layer of quoting).

First seen in :

## User


```python
# tool result (toolu_bdrk_01CmkqJNgwQLAG4HseMnvAR6)
"{"email":"[email protected]"}"
```

Desired output:

## User


```python
# tool result (toolu_bdrk_01CmkqJNgwQLAG4HseMnvAR6)
{"email":"[email protected]"}
```

---

When using repr(), strings will be quoted. JSON will be given as is.

>>> a = str("asfd") 
>>> print(f"\n{a}\n")

asfd

>>> a = repr("asfd")
>>> 
>>> print(f"\n{a}\n")

'asfd'

>>> a = str({"a": 1, "b": 2})
>>> print(f"\n{a}\n")

{'a': 1, 'b': 2}

>>> a = repr({"a": 1, "b": 2})
>>> print(f"\n{a}\n")

{'a': 1, 'b': 2}

[schloerke]

Still isn't what I want

[schloerke]

Need to use pprint.pformat, but I can see rich is already available.

[schloerke]

😂 Link

E       assert 'Joe: sage green' in "I cannot answer this question. The available function `favorite_color` does not contain any information about Joe and Hadley's favorite colors.\n"
E        +  where "I cannot answer this question. The available function `favorite_color` does not contain any information about Joe and Hadley's favorite colors.\n" = str(<chatlas._chat.ChatResponse object at 0x7fcc73020a70>)

[schloerke]

The test that is failing is a false-positive. I re-ran it once and it did not improve. 🙄

[schloerke]

Got it to pass by waiting. woot