You'd be surprised how common this question is, and how few answers there are. Colored gases tend to be corrosive and poisonous, such as NO2, I2 and Br2 vapor. Sometimes you can use a dense gas such as SF6, shine a beam of light under the device being tested onto a white screen, and look for the density waves on the screen. Again, this takes a fairly big leak to get a response.
The traditional way of detecting gas leaks is to pick up the sound. Sometimes your ear is okay for this, but for the small scale you're talking about, ultrasonic methods are more sensitive. You might be able to do it on the cheap, with a crystal microphone with response into the 40 KHz, an amplifier, and an oscilloscope. Sometimes the sounds get up into the MHz and it gets very expensive.
Do you have to locate the leak or just determine that there is a leak? Sometimes a change in weight over time will do the latter, especially if you fill with a heavy gas such as SF6 or dimethyl ether.
Finally, you can build a little vacuum system with a fine tip and a flow rate of about 10-20 cc/min. Draw the gas over a carbon monoxide sensor. Fill the device with CO and use the sampling tip like a geiger counter to locate the leak. Building this would cost you $50 to $400 depending on the sensor and the vacuum system you use.
Dangerous Bill